Quantitative Diatom Assemblage Data 1. Sample Preparation

Scotia Sea cores TPC063 and TPC286 were initially sampled at intervals of 32 cm, and at higher resolution (4 cm) in specific portions of interest referred to as the high resolution windows (HRWs). The upper 380 cm of core MD03-2603 was sampled at a resolution of 12 cm, with a HRW, contemporaneous with those in the Scotia Sea cores, sampled at higher resolution intervals of 4 cm.

Sediment samples were dried and weighed in preparation for chemical treatment. Sediment mass is dependent upon the sediment composition and the relative contribution of terrigenous and biogenic material. South o f the APF, glacial sediments are characterised by higher terrigenous input than biogenic input, which results in more terrigenous grains than diatom valves per gram of sediment and the necessity for samples with a higher mass in order to achieve suitable diatom content. A high proportion o f terrigenous grains can overwhelm diatom valve presence making analysis of samples difficult. Samples of a mass adequate to produce clear quantitative slides were therefore estimated using a combination of the MS record, to infer the likely concentrations of terrigenous grains, and smear slides to assess the respective proportions o f clay grains and diatom valves. Subsequent to weighing, the sediment was digested in a solution of hydrochloric acid, hydrogen peroxide and Calgon™ in order to remove carbonates, dissolve organic detritus and to further disaggregate the sample.

Quantitative slide preparation followed the settling method o f Scherer (1994), whereby each sample is settled through a 1 litre beaker of water over a period of 4 hours to minimise sediment clumping and maximise even distribution. After 4 hours the water was drip-drained overnight, allowing the cover slips to dry in order to be mounted upon a permanent microscope slide using Norland Optical Adhesive 61 (refractive index 1.56).

3.1.4.2. Assemblage Data 3.1.4.2.1. Valve Counts

Glacial sediments are often characterised by reduced concentrations of diatoms and enhanced quantities o f terrigenous detritus. In addition to this, the diatoms that are preserved within the sediments can often be lightly silicifed and heavily fragmented. Fragmentation creates a risk

of counting duplicate valves and thus over-estimating species abundance within an individual sample. This was avoided by following the strategy employed by Armand (1997), Cunningham and Leventer (1998) and Allen (2003). In all samples diatoms were identified to species level. Where detailed identification to species level was not possible valves were assigned to genus or included as an unidentified species. The original count sheets o f diatom species assemblages can be viewed in Appendix 1.

Valve counts were converted into both relative and absolute abundances (as described by Scherer (1994), which allows the calculation of absolute diatom concentration per gram of sediment), reducing the species bias problems that can sometimes affect relative abundance datasets and eradicating the ‘closure issue’ also associated with relative abundance data.

Absolute abundances, for both total diatom valve concentration and individual species were calculated using the following equation (Scherer, 1994):

T = ((NB)/(AF)) M

N = Number o f diatoms counted

B = Area o f beaker bottom (7854 mm2) A = Area of transect

F = Number of transects M = Mass of sediment (g)

3.1.4.2.2.Diatom Counting Methods

Two microscope-based diatom valve counting methods were employed.

3.1.4.2.2.1. Conventional Counting Method

Where possible, between 300 and 450 valves (in accordance with the literature) were counted along slide transects using an Olympus BH-2 microscope at xlOOO magnification (Zielinski and Gersonde, 1997; Cunningham and Leventer, 1998; Gersonde and Zielinski, 2000;

Sjunneskog and Taylor, 2002; Allen et al.y 2005; Buffen et al., 2007).

3.1.4.2.2.2. Area Counting Method

In addition to the conventional counting method a second method, more widely employed in macro-ecology quadrat counting (Fortin and Dale, 2008) was also used. This is based on the quantification of species within a statistically significant area of the quadrat/slide coverslip.

Taxonomic identification was undertaken using an Olympus BH-2 microscope xlOOO

magnification for a statistically significant portion (10 transects) of the 22 x 22 mm2 coverslip (Buffen et al., 2007). This area-based method was necessary as a number of the samples from core TPC286 were barren of diatom valves and precluded use o f the conventional counting method as counts of between 300 and 450 valves were not possible. These samples, in core TPC286, are primarily located between 406 cm and 543 cm, and 350 and 238 cm representing a major portion of the record (Figure 3.9a). Alternative attempts made to amplify the quantity of diatom valves per cover slip through increasing the sediment mass of each sample obscured the frustules with terrigenous grains.

a) Total Diatom Concentration (millions of v/gds)

b)

Hyalochaete Chaetoceros resting spores Relative Abundance (%)

(exc. dummy variable) 0 20 40 60 80 100

c)

Hyalochaete Chaetoceros resting spores Relative Abundance (%)

(inc. dummy variable) 0 20 40 60 80 100

Period of low or no diatom valve deposition

Figure 3.9. a) Core TPC286: down core total diatom concentration highlighting periods o f low or no diatom deposition, b) Core TPC286: down core relative abundance o f Chaetoceros rs calculated without the benefit o f the dummy variable, c) Core TPC286: down core relative abundance of Chaetoceros rs calculated with the benefit o f the dummy variable. Grey bands indicate periods o f low or no diatom valve deposition.

The use of this area counting method requires caution when calculating relative abundance due to the exceptionally small sample sizes (in some cases as few as one solitary diatom valve) (Figure 3.9b). In order to compensate for this bias a measure of diatom absence was

introduced to the data in the form o f a ‘dummy variable’• For any sample showing diatom depletion, where a tally o f 300 to 450 values was impossible, the ‘dummy variable’ was used to make up the difference, and was given the value required to reach a significant tally of 400 valves, i.e. for a sample that only identifies 10 diatom valves, the ‘dummy variable’ was allocated a value of 390. This variable essentially represents the portion of the coverslip that is not diatoms, giving us a means to measure low rates o f diatom deposition within the relative abundance statistics, which increases the comparability of relative abundance data throughout the core and with other cores (Figure 3.9c).