The sampling programme was conducted at a fixed station R1 (66°39.223′S, 140° 00.139′ E; Fig. 1) at a water depth of 53 m, ca. 300 m off the Dumont d'Urville Station on Petrel Island (Île des Pétrels) in East Antarctica, from 1 November (day 1) to 29 December 2001 (day 59). The station R1 was the same one visited in previous expeditions (Riaux-Gobin et al. 2000, Riaux-Gobin et al. 2003, Riaux-Gobin et al. 2005). The ice conditions and structure were the same as described by Riaux-Gobin et al. (2003), with the ice break-up occurring on 10 January 2002 in this study. Discrete daily sampling was performed on the Bi, Pi and under-ice water at 5 m (UIW5).

Infrared satellite images (Fig. 2) and meteorological data (Fig. 3) were provided by the Dumont d'Urville meteorological station (Météo France and MétéoSat). The satellite images were obtained from NOAA12 and NOAA14 satellites, equipped with an advanced very high-resolution radiometer with five channels and 1-km vertical resolution and 4-km lateral resolution. After computer-aided processing provided by the Service d'Archivage et de Traitement Météorologique des Observations Satellitaires, in Lannion, France, the composite images were used to distinguish ice, water and cloud cover. The continental boundary is taken from map 6061, provided by the Service Hydrographique et Océanographique de la Marine. The limits of the glacier tongues—particularly the highly variable Dibble Tongue are artificially included as “continental limits” (Fig. 2). The day length measured by a heliograph with fibre optics at the meteorological station at Dumont d'Urville Station (Fig. 1) is defined as the time during which the light level was >120 W m⁻² (Fig. 3). The theoretical tidal data (Fig. 3) were obtained from the French Naval Hydrographic and Oceanographic Service (Pineau 1999; Le Roy & Simon 2003) and the Subantarctic and Antarctic Sea level Observation Network, established by the Laboratory for Studies in Geophysics and Spatial Oceanography in Toulouse.

Two multiple MST-6 sediment traps (Hydro-Bios, Kiel, Germany) were deployed under the land-fast ice at R1 for five weeks from 8 November to 6 December (Fig. 4). Each trap had an opening area 0.015 m², covered by a 40-mm thick plastic grid, with 40 mm² holes. Each trap was equipped with six 250-ml sample cups, filled with filtered seawater poisoned with 1% HgCl₂ that rotated at a pre-set interval of six days. The top trap was located at 3 m under the land-fast ice, with an RCM7 current meter (Aanderaa Data Instruments, Nesttun, Norway) directly underneath it, while the bottom trap was positioned at 6 m above the seafloor (Fig. 3). The whole array system was stabilized with a mooring anchor. After the recovery of the sediment traps, the sample cups were split into sub-samples for later determinations of POC, nitrogen (PON) and chloropigments. The current meter, which was equipped with a temperature sensor, was programmed to register data every 20 min. The progressive diagram of currents was calculated with Aanderaa software.

Sub-ice salinity and temperature profiles (0–50 m) were obtained from 3 November to 14 December using an SBE 37-SI MicroCAT conductivity–temperature profiler equipped with an SBE ST pump (Sea-Bird Electronics, Bellevue, WA, USA). The instrument had a resolution of 0.0001°C and 0.0001 psu and was calibrated by Sea-Bird Electronics prior to and after use. Daily profiles were always taken at noon.

During the 59-day sampling programme, a total of 19 discrete samplings of Bi, Pi and underlying water at 5 m water depth (UWI5) were acquired from 1 November (day 1) to 29 December (day 59) within a reduced surface (<60 m²) to minimize the impact of spatial heterogeneity/patchiness (see Discussion section).

On each sampling day, snow and ice thickness and sub-ice irradiance were measured at noon local time. The instrument used was an SA light sensor (Li-COR Biosciences, Lincoln, NE, USA), equipped with a Spherical Quantum Sensor LI-193SB, operating through the ice coring hole (around 8 cm in diameter), which was covered by a lid. The resulting 0–40 m profiles give a rough estimate of the light reaching the sympagic algal layer at the base of the ice column.

Two to three ice cores were collected during each visit with a SIPRE coring auger (i.e., of the type developed by the US Army Snow, Ice and Permafrost Research Establishment) with an inner diameter of 7.5 cm. Each core was stored separately in a black plastic bag and kept at −20°C at the Dumont d'Urville Station where the last 5 cm of the Bi was removed with a stainless steel saw. The sherbet-like Pi was rapidly scooped out from the core hole with a stainless steel ladle, stored in plastic or glass containers and kept at −20°C. As a rough indication, the Pi thickness was evaluated when scooped. Ice core samples were slowly melted in the dark at 4–7°C from a few hours for Pi samples to 12–16 h for Bi samples. Water samples were collected at 5 m depth (UIW5) with a simple hydrological bottle consisting of a weighted polyethylene cylinder with a stopper operating from the surface (Riaux-Gobin et al. 2003).

The consolidated land-fast ice at station R1 off Pierre Lejay Bay had a maximum thickness of 1.40 m measured in early November, with a thin Pi layer that was on average 0.20 m thick. The Bi horizon consists of granular ice colonized by sympagic algae over a thickness of 1–5 cm. The Pi thickness varied from 0.20 m at the coast to 0.30–0.40 m at 6-km offshore (data not shown). The land-fast ice was in place until the end of November 2001 and extended on average 80 km in front of the Dumont d'Urville Station (Fig. 2a). A large ice-free area was observed between the pack and land-fast ice until 22 November, when strong katabatic winds from the continent enlarged this polynya. The ice break-up started 30–40 km offshore on 4 December (Fig. 2b, c). On 1 January 2002 (Fig. 2d), the ice break-up was visible along a large part of the coast. The final dislocation of the 1 m thick land-fast ice in Pierre Lejay Bay occurred on 10 January under a noticeable swell (pers. obs.; no swell recorder) and strong wind from the east.

The fast-ice thickness at R1 (Fig. 5) slowly decreased from 1.32 m on 3–5 November to 1.00 m on 29 December. The snow cover decreased from 0.10 to 0.00 m (Fig. 5), albeit with several brief increases due to new precipitation (i.e., around days 36 and 50). The ice surface became soft and porous by 27 December but without melt ponds. The Bi horizon also softened and became unstable at that time while the Pi remained present at R1 with around the same thickness (0.20 m±0.1) during the entire study period. The Pi was deeply coloured by sympagic algae by 23 November. After 29 December, the ice began to rot throughout, making travel on the fast ice unsafe, so sampling was discontinued.

The mean air temperature progressively increased from −14°C in early November 2001 to above freezing point in late December (Fig. 3). Six consecutive days with positive temperatures characterized the time between the end of December and the beginning of January. A new record in maximum air temperature (9.9°C) was measured on 30 December 2001.

The tides in Pierre Lejay Bay are composed of two irregular daily oscillations (Fig. 3). During the study period, the maximum tidal amplitude was 2.2 m and the maximum current speed was 35 cm s⁻¹. Maximum current speed events were associated with the mid-flood and mid-ebb periods of the maximal daily oscillation (data not shown). The coastal surface currents were directly under the influence of the tidal regime. The reverse between ebb and flood was sudden, with a short but strong northern component just before the flood and a high reproducibility between consecutive tidal cycles (data not shown).

The progressive vector of currents (Fig. 7) showed two major components: east–north-east (from A to B and from C to D) and west–north-west (from B to C and from D to E); these two major directions of the drift were related to neap tides (east–north-east drift) and spring tides (west–north-west drift). Further registrations would help in understanding the exact parameters forcing the changes in drift directions: B, C and D in Fig. 7 are roughly located at the start and the end of the spring tide period.

Average and range values of physical, chemical and biological variables from Bi, Pi and UIW5 at R1 during the study period are summarized in Table 1. Temporal variations in salinity, Si(OH)₄, N O 2 - , N O 3 - + N O 2 - , P O 4 - , Chl a, Phaeo a and pH before the ice break-up are illustrated in Figs. 8 and 9.

A steady decrease in salinity was obvious in all layers (Bi, Pi) and UIW5 during the days preceding the ice break-up (days 47–59; Fig. 8a), even if less marked for the Bi. After day 30, the Chl a biomass in the Bi increased (Fig. 9b), but with no regular trend, and reached 2.6 mg l⁻¹ (i.e., 130 mg m⁻²). The Chl a concentration was on average 0.6 mg l⁻¹ during the 70 days preceding the ice break-up (i.e., 30 mg m⁻², Table 1). The POC concentration averaged 20 mg l⁻¹ (i.e., 1 g m⁻²; data not shown) in the Bi horizon. These values were positively related from day 35 to an increase in under-ice irradiance until the ice break-up (Fig. 5b). Note the great temporal variations in the ice algal biomass (Fig. 9) and in the succession of the different microalgal taxa (Fig. 11–13).

P O 4 - concentrations, reaching a maximum of 51 µM in Bi and 38 µM in Pi at day 55, showed a co-variation with Chl a (Fig. 9a, b). High silicic acid concentrations of >70 µM at 5-m water depth at the end of the time series, and at 10 and 20 m in the underlying water masses (data not shown) characterized the sub-ice water while low values were recorded in the ice (particularly the Bi, i.e., 5–7 µM), suggesting a possible biological depletion (Fig. 8b). The Si(OH)₄ concentrations at 5 m clearly increased from day 30 (Fig. 8b), whereas N O 3 - + N O 2 - concentrations were high throughout the study, particularly at the end of the time series, where it reached 67 µM in the Pi horizon. Nitrite (Fig. 8c) was related to pigment degradation in Bi and Pi horizons before the ice break-up when Phaeo a reached 850 µg l⁻¹ at the end of the time series (Fig. 9c). The pH values ranged from 7.2 to 8.4, without any specific trend, except for slightly lower values in the Bi horizon towards the end of the time series (Fig. 9d). Graphically represented in an MDS ordination, the environmental variables (nutrients, pigments, salinity, pH) clearly discriminated three groups of samples consisting of Bi, Pi and UIW5 (Fig. 10). This analysis confirmed the probable distinctive origin of Bi and Pi, and supports the observed differences in time variation in the three horizons.

POC and PON concentrations in the Bi and Pi horizons varied from 15 to 21 mg l⁻¹ and from 2 to 3 mg l⁻¹, respectively. They are 10–20 times higher than in the sub-ice waters (Table 1). The C:N ratio, averaging around 8 in the ice horizons was lower than in UIW5 (14.2).

The POM flux at R1 remained stable during the sediment trap experiment although there was a sharp increase in the last cup, 33 days before the effective ice break-up (Table 2). In the near-surface trap at 3 m under the ice (Fig. 4), the pigment concentrations increased slowly and more or less regularly from 8 November (day 10), but more strongly after 2 December (day 34), with a concentration 10 times higher than at the beginning of the trap experiment. In the bottom trap, at 6 m above the seafloor, all pigment concentrations were higher than those measured directly under the ice and here again after 2 December the pigment concentration was six times higher than at day 10. The Chl a:Phaeo a ratio was surprisingly high (>0.9) and higher in the bottom (1.5–2.0) than in the surface trap (0.9–1.3). It did not significantly vary with time. The Chl c:Chl a ratio varied similarly with time on the two traps (Table 2) and its mean value (on average 0.19) was not so different than in the Bi (on average 0.21) and Pi (on average 0.22) sympagic communities (Table 1, ratio not detailed). POC and PON also increased in the two sediment traps with maximum values of 62 and 16 mg m⁻² d⁻¹, respectively, in the surface layer and slightly higher rates at 47-m depth (79 and 24 mg m⁻² d⁻¹). The C:N ratio was very low (2.4–5.0, Table 2) in comparison to the values recorded in sea ice or UIW5 (8.3–14.2, Table 1) and with the more classical Redfield ratio in phytoplankton (6.6; see Redfield et al. 1963).

The trap material contained few zooplankton, attesting to a possible low sub-ice level of grazing near the continent shelf, as previously recorded by Honjo (2004) in the Ross Sea Gyre. A mini-remotely operated vehicle (ROV) survey under the land-fast ice near R1 in 2007 found no krill (G.S. Dieckmann, pers. obs.), supporting the suggestion of a low grazing pressure in this area, irrespective of the sampling year.

From the maximum particle flux at R1 (Table 2), it is difficult to extrapolate to an annual flux value since the duration of the most intense events, that started several weeks before the ice break-up at coast, is unknown. Nevertheless, we can consider that in 2001/02, the total rate in Pierre Lejay Bay must have been ≥7 g POC m⁻² yr⁻¹, with a potential active sedimentation period of three months around the ice break-up (see Discussion section).

Microalgae recorded in the Bi, Pi and UIW5 are presented in Table 3 while the successions of major protist groups and species can be viewed in Figs. 11–13. The similarities and dissimilarities (SIMPER) between Bi, Pi and at UIW5 are presented in Table and 5 .

Four diatom taxa, namely Berkeleya adeliensis, Cylindrotheca closterium, Nitzschia lecointei and N. stellata, were preferentially associated with the Bi horizon reaching abundances of 21×10⁸ cells m⁻², 90×10⁶ cells m⁻², 19.5×10⁷ cells m⁻² and 17.5×10⁸ cells m⁻², respectively, while Navicula glaciei, with 49×10⁸ cells m⁻², and Amphiprora kufferathii, with 15×10⁶ cells m⁻², were better adapted to the Pi horizon (Figs. 11, 12). Berkeleya adeliensis and Fragilariopsis cylindrus first occurred in the Bi with 16×10⁷ cells m⁻² for F. cylindrus, but later colonized the Pi horizon (Fig. 12). The same differentiation in colonization between Bi and Pi can be observed for the nanoflagellates (Fig. 13).

Interestingly, centric diatoms occurred more frequently in the Pi than Bi horizon (Table 3), while the pennate genera Haslea Simonsen and Plagiotropis Pfitzer were more common in the Bi horizon and flagellates from the UIW5 depth. Furthermore, several algal taxa were only present in the UIW5, such as some Dinophyceae, Entomoneis oestrupii, Fragilariopsis obliquecostata and Pyramimonas nansenii (Table 3). Gymnodinioid taxa were encountered in sub-ice waters, probably developing a heterotrophic strategy or specific adaptation to low light, while the archaeomonads were present in all sea-ice horizons but far less abundant at 5 m depth.

The reported diatom diversity was relatively low, in part due to the methodology. The Shannon (H’) index averaged 1.63 in the Bi, 1.60 in Pi and 1.89 at UIW5. The H’ index slightly increased over time in the Pi horizon and UIW5 (data not shown).

A succession in the sympagic communities was obvious for diatoms and other microalgal taxa (Figs. 11–13). Among the dominant species, Amphiprora kufferathii, Cylindrotheca closterium, Navicula glaciei and Nitzschia lecointei reached maximum cell abundance during the first half of November (Fig. 11), while Fragilariopsis cylindrus, F. sublinearis, with 28×10⁶ cells m⁻², and Nitzschia stellata, with 92×10⁸ cells m⁻², reached their higher abundance in the second half of that month (Fig. 12). Other taxa, such as Berkeleya adeliensis and Fragilariopsis curta, with 35×10⁵ cells m⁻², were dominant for a few weeks before the ice break-up. Nanoflagellates were more abundant towards the end of November in the Bi horizon, reaching 26×10⁸ cells m⁻², while gymnodinioid dinoflagellates (<20 µm) were increasing in number in the Bi by mid-December (Fig. 13). The prymnesiophyte Phaeocystis antarctica, reaching 3.6×10⁶ cells m⁻², grew preferentially in the Pi and at UIW5, but only during the last 10 days before the ice break-up (Fig. 13d).

The clustering analysis (Fig. 14) discriminated two groups, one composed of Phaeocystis antarctica, Fragilariopsis curta and gymnodinioid cells (<20 µm), which were characteristic of the UIW5 depth, and one composed of 10 other diatom taxa and nanoflagellates (<10 µm).

The centric and pennate empty frustules were not included in the analyses but 54% of the total diatom population at UIW5 consisted of empty cells, with maximum values from day 20 to day 50. This percentage was lower but extremely variable (from 1% to exceptionally 40 %) in the Bi and Pi horizons and increased slightly until the ice break-up, but with no direct relationships with the degraded pigment concentrations (Phaeo a).

Some correlation analyses between microalgal species and different subsets of environmental variables (BIOENV analysis) showed that Chl a and c concentrations were highly correlated to the Phaeo b concentrations, and between themselves (R>0.90, p<0.001 in all cases), and were therefore excluded from the analysis (Clarke & Warwick 2001). The structure of the microalgal communities was explained by the best three-variable combination (R=0.68), corresponding to salinity, Phaeo b and P O 4 - concentration.

In the vicinity of the Dumont d'Urville Station, the land-fast ice thickness was relatively consistent between years (Riaux-Gobin et al. 2000, Riaux-Gobin et al. 2005). The spatial heterogeneity was tested at R1 in 1999 while the fast-ice conditions were similar to that experienced in 2001 (Riaux-Gobin et al. 2000, Riaux-Gobin et al. 2005). The test took place on a large surface (average 625 m²), following a “Latin Square” pattern (Table 6). Each of the five coring points was sampled twice (duplicates at 3-m intervals). Data were not homogeneous across the sampled area, which was probably related to the differences in snow cover or irregularity in ice thickness, while the duplicates showed a lower variability (Table 6). To reduce the impact of the patchiness, the 2001 time series took place on a reduced area (see Sampling site, material and methods section).

During the austral spring of 2001, the biochemical conditions of the land-fast ice prior to the break-up were relatively similar to those reported from the same Dumont d'Urville sampling site in 1995 and 1999 (Riaux-Gobin et al. 2005). The temporal variation of the nutrient and pigment concentrations matched those previously recorded (Riaux-Gobin et al. 2000, Riaux-Gobin et al. 2005). However, there were some differences that can be linked to the year's conditions. For instance, the absence of early winter break-up events in 2001 may have caused the exceptionally high thickness of the ice and delayed the timing of the break-up in early 2002 when compared with the timing in 1995 and 1999 (Riaux-Gobin et al. 2005).

The 2001 nutrient concentrations were lower than those recorded at the same station R1 in 1995 (300 µM N O 3 - in Pi) but similar to 1999 (Riaux-Gobin et al. 2005). Fiala et al. (2006) reported much lower nutrient and algal biomass concentrations in 1998 at station C, near to R1, with 10–12 µM N O 3 - and 50–90 µg Chl a l⁻¹ in the Bi horizon, highlighting the interannual variability in the measured parameters. The decoupling between the silicon cycle and the nitrogen cycle (Riaux-Gobin et al. 2000, Riaux-Gobin et al. 2005) was confirmed by the present data set. A decoupling between the silicon and the phosphorus cycles (Figs. 5, 7) is also highlighted. High P O 4 - concentration in the Bi horizon reaching 51 µM at the end of this study is indicative of a mineralization of the organic matter (Arrigo et al. 1995; Günther et al. 1999). Nitrification concurrently took place in Bi and Pi while the under-ice water was enriched with N O 2 - at the end of the study. Si(OH₄) was depleted in the 2001 ice horizons and showed high concentrations at 5 m depth, corroborating previous reports (Riaux-Gobin et al. 2000, Riaux-Gobin et al. 2005).

Some interannual differences appeared in the microalgal colonization of the various ice horizons, which may be explained by their formation history. The maximum Bi Chl a biomass (130 mg m⁻²) recorded in 2001 was close to that reported in 1999 (87–187 mg m⁻²; Riaux-Gobin et al. 2005). Conversely, the maximum Chl a biomass in 1995—1.9 g m⁻²—was considerably higher but was recorded in the Pi horizon (Riaux-Gobin et al. 2005). This biomass difference between years can be partly explained by difference in ice thickness and its effect on light availability. Ice thickness was low (60 cm) at the 1995 break-up compared to that in 2001 (101 cm). The timing of the spring break-up may also be related to the ice thickness and to biomass differences between years: the break-up occurred between 21 and 27 December in 1995 compared to 10 January 2002. The high biomass may generate heat that could accelerate the melting of the ice. The microalgal biomass increased, but irregularly, in Bi and Pi in 2001, with several peaks, especially during the last period of time series (Fig. 9b). A noticeable increase in the sub-ice irradiance at 1 m was positively related to the snow cover ablation (Fig. 5), but with several lower records in relation to snow events. The snow cover is an important factor that decreases the light penetration to the Bi and, therefore, may influence the variations in the sympagic microalgal biomass (Arrigo et al. 1995). There are three periods showing a significant increase in Chl a biomass in 2001 (around day 20 and day 32 and between days 47 and 60; Fig. 5) that may be linked, with a lag of 5–10 days, to an increase in irradiance and snow cover ablation (Fig. 5). The Phaeo a concentrations in Bi and Pi strikingly increased towards the last 10 days of the 2001 sampling period, in parallel with a decrease of the Chl a:Phaeo a ratio (data not shown), which can be explained by a decaying Bi community and unhealthy algal cells. This is confirmed by high P O 4 - and N O 2 - concentrations in Bi and Pi at the end of the study period, which was related to remineralization of the organic material.

The high proportion of empty frustules in the sub-ice waters in 2001 may signal active sedimentation from the productive Bi and Pi horizons and a fast turnover of the microalgal organic matter through the bacterial loop or grazing long before the ice break-up (Lizotte 2003). Nevertheless, copepods and other metazoa were rare in the land-fast ice off Adélie Land (Fiala et al. 2006; G.S. Dieckmann, pers. obs.). The function of the grazing in the ice is therefore unclear. The microbial loop activity at the sea-ice interface is also unclear since the bacterial biomass is assumed to be low in the land-fast ice off Adélie Land: bacteria and protozoa represent only 16.4 % and 3% of the total carbon microbial biomass, respectively (Delille et al. 2002).

The particle flux under the fast ice has been reported to sharply increase during a short period of two to three months in Lützow-Holm Bay (Matsuda et al. 1987; Leventer 2003). The extrapolate fast-ice flux off Adélie Land (≥7 g POC m⁻² yr⁻¹) is higher than that reported from the Weddell Sea in similar conditions (Fischer et al. 1988: 0.37 g m⁻² yr⁻¹), but lower than from Bransfield Strait or the Ross Sea (Wefer et al. 1990: 60 g m⁻² yr⁻¹; Leventer 2003). Regarding the high variability of these under-ice fluxes around Antarctica, we can note that the Pi horizon is variable in thickness and may restrict biogeochemical exchanges between the sea ice and the underlying waters (Thomas et al. 2001).

The relatively high pigment concentrations of the sedimented material at spring (as also observed in Lützow Bay by Matsuda et al. [1987]) argue for a fast downward particle flux near the coast. Nevertheless, the sub-ice “progressive vector of currents” in surface waters off the Dumont d'Urville Station, recorded for the first time in that area, supported an effective offshore export of the particulate matter. Further experiments would allow the quantification of the particulate matter potentially dispersed offshore by this drift. The preservation of microalgae and the downward fluxes to the sediment surface have been reported in other parts of Antarctica—for example, in Bransfield Strait (Kim et al. 2004) and in Lützow-Holm Bay (Tanimura et al. 1990)—and may contribute to palaeo-reconstructions (Armand & Leventer 2003; Buffen et al. 2007). Furthermore, this progressive movement of currents oriented west–north-west and the open shape of Pierre Lejay Bay would support an outflow of the ice pack after the break-up and may also partially explain why the land-fast ice is relatively thin in this part of Antarctica.

As reported in Ushio (2006) for Lützow-Holm Bay, the fast-ice break-up conditions in the Antarctic vary considerably from one year to another, and depend on oceanic and atmospheric conditions, which are rarely discussed in ecological surveys. In 2001, the ice break-up in the Pierre Lejay Bay was delayed by approximately two weeks compared to average annual data (not shown) and occurred on 10 January 2002 (day 70). Several periods with strong winds coming from the continent (up to 130 km h⁻¹) in early January 2002 (Fig. 3) may trigger some upwelling events of warmer deep waters, as suggested in Fig. 6. During this same period, a steady and significant swell along with strong east wind contributed to the ice break-up along the coast, even if the ice was still 1 m thick. The effective ice break-up at the coast on Pierre Lejay Bay occurred one month after the end of the study reported in this article. However, the pack ice was breaking up 100–150 km offshore some weeks before the experiment, as observed on satellite imagery (Fig. 2) and as generally observed at that period of the year in the area of the station (Riaux-Gobin et al. 2005).

As reported by Krebs et al. (1987) and Lizotte (2001), the temporal variation in species composition of first-year land-fast ice and whether these changes are related to sea-ice ecology are scarce for the Antarctic (Cota & Sullivan 1990; Watanabe et al. 1990; Gleitz et al. 1998). Moline & Prézelin (1996) reported factors regulating the variability in coastal Antarctic phytoplankton and they stated that the changes in phytoplankton successional patterns may be a more sensitive marker of detecting long-term trends in Southern Ocean ecosystems than biomass or productivity indices. The succession of sympagic microalgae may be a good marker of environmental sea-ice changes, while the microalgal biomass is variable from one year to another, as reported in this study. The 2001 sympagic microalgal time succession was the first to be analysed and illustrated for the Dumont d'Urville area. Drawing from results of previous studies at the same sampling site, we can point out that the dominant species were the same (Riaux-Gobin et al. 2005). Nevertheless, in 1995 (Riaux-Gobin et al. 2005) the colonization of the different ice horizons may have been slightly different, with the Pi more deeply colonized than in 2001 when the Pi was enriched in microalgae only at the end of the sampling period. It would be of interest to study other time series in the same area, with comparable environmental conditions, to arrive at a conclusion regarding the reproducibility of the taxa succession.

In 2001, a succession took place with some taxa appearing late in the season such as Berkeleya adeliensis and Phaeocystis antarctica (Figs. 12, 13). Some taxonomic differences appeared between the Bi and Pi horizons throughout the sampling period (Table , 5 , Figs. 11–13). For example, Fragilariopsis cylindrus and Nitzschia stellata were mostly associated with the Bi while Amphiprora kufferathii and Navicula glaciei mainly occurred in the Pi. These observations support the hypothesis about the different origin—history and way of formation—of the Bi and Pi layers, as first reported by Riaux-Gobin et al. (2000). Similar taxonomic differences between Bi and Pi were also recorded in the Weddell Sea by Günther & Dieckmann (2001). Some species switched over time from the Bi to the Pi, such as Berkeleya adeliensis, Fragilariopsis curta, F. cylindrus and Nitzschia lecointei. The opportunistic tube-dwelling sea-ice diatom B. adeliensis is apparently well adapted to changes in texture during ice melting, while other diatoms died or were flushed from the Bi to the Pi and the water column, probably due to the physical changes occurring in the melting ice. An increase in B. adeliensis during melting was also noticed in 1995 off Adélie Land (Riaux-Gobin & Poulin 2004). This Antarctic ice diatom is known to form tubular colonies creating mats or hanging strands underneath the bottom part of the ice (Hoshiai 1977; Sasaki & Hoshiai 1986; Watanabe 1988; McMinn et al. 2000).

Amphiprora kufferathii, nearly absent at the end of the present time series, was also noticed to decrease in late December in the Bi near Syowa Station, whereas Nitzschia stellata was increasing concurrently (Watanabe et al. 1990). Archer et al. (1996), describing a fast-ice community near Davis Station, showed the decay of autotrophic diatoms such as Entomoneis Ehrenberg, followed by an increase in heterotrophic organisms (i.e., euglenoids and dinoflagellates). In the present study, nanoflagellates and small gymnodinioid cells also appeared in the Bi after the disappearance of A. kufferathii, Navicula glaciei and Nitzschia lecointei.

The last weeks before the ice break-up are characterized by: (1) a higher under-ice irradiance that triggers a sympagic production; (2) the ice melt, promoting the growth of flagellates and particularly prymnesiophytes; and (3) a peak of degraded pigments and the decay of a large part of the colonial pennate diatom community such as Amphiprora kufferathii or Berkeleya adeliensis that form macroscopic clumps, noticed loosely attached to the bottom of the Bi at the end of the time series.