Sensitivity test on the effect of sea surface temperatures on iceberg melting patterns

To better understand Pliocene IRD provenance patterns in the context of variable climatological conditions, constraints must be placed on the relative importance of iceberg transport pathways (see Section 6.4), the sediment-loading of icebergs, iceberg flux to the Southern Ocean, and iceberg survivability. IRD deposition is ultimately a complex function of all of these processes. While sediment loading and iceberg fluxes are intrinsically difficult to constrain, iceberg survivability is largely a function of iceberg volume, current speed and melting rates, which in turn is controlled by prevailing SSTs.

5.8.1 Iceberg trajectory modelling framework, and comparison of results from modelled pre-industrial iceberg melting patterns and observed core-top IRD provenance patterns

In order to investigate the role of SSTs on iceberg trajectories and to allow for comparison to observed Pliocene IRD patterns, an iceberg survivability model developed with colleagues Daniel Hill (University of Leeds, BGS), Aisling Dolan, and Alan Haywood

Figure 5.8. Histogram illustrating the data-model comparison for Wilkes Land IRD relative to Prydz Bay IRD deposited (light grey boxes), and Wilkes Land icebergs relative to Prydz Bay icebergs melted (dark grey boxes) at four sites located in the Prydz Bay region during the Holocene. Observed IRD provenance is inferred from ⁴⁰Ar/³⁹Ar age distributions of hornblende grains (Brachfeld et al, 2007; Roy et al. 2007; this study).

(University of Leeds), following previous work by Gladstone et al. (2001) and Bigg et al.

(1997). After Gladstone et al. (2001), the model injects 100 icebergs of 10 different size classes (ranging from 40x40x60m to 250x1467x2200m in volume) into the Southern Ocean from seven distinct drainage basins located in the Prydz Bay area (represented by the Lambert, Philippi and Denman Glaciers) and at the Wilkes Land margin (represented by the Adams, Totten, Thompson and Frost Glaciers) (see Figure 5.1 for locations of drainage basins), and subsequently records their trajectories and location of melting. Icebergs were released monthly over a 12 month period (i.e. 1200 from each glacier in total) with seasonally changing wind fields and surface currents taken from a pre-industrial global climate model scenario (HadCM3). Annually averaged modelled iceberg trajectories derived from this model set-up match well with those observed by satellites today (Antarctic Iceberg Tracking Database [1978–2012]; available at http://www.scp.byu.edu/data/iceberg/ database1.html), with icebergs travelling in an anti-clockwise direction upon entrainment into the westward-flowing coastal current (Figure 5.2).

To evaluate this iceberg trajectory modelling framework, iceberg melting patterns were modelled over five core sites (ELT47-07, ELT47-14, ODP Site 1166, ODP Site 1165, and RC17-51; core-top data from Brachfeld et al. 2007; Roy et al. 2007; this study; Figure 5.8), all located in the Prydz Bay area, under pre-industrial climate conditions. Importantly, modelled icebergs did not contain IRD, rather their site of melting was modelled, with each

Figure 5.9. a) Percentage of icebergs produced from Prydz Bay glaciers (in orange) and Wilkes Land glaciers (in blue) that melt over ODP Site 1165 in different SST scenarios. b) Calving flux (Gladstone et al. 2001) and approximate distance from ODP Site 1165 of Prydz Bay glaciers (orange data points) and Wilkes Land glaciers (blue data points). c) Combined percentage of icebergs from Prydz Bay (orange line) and Wilkes Land (blue line) glaciers (left y-axis) that melt at ODP Site 1165 ratio under different SSTs (x-axis). The modelled ratio of Wilkes Land icebergs to Prydz Bay icebergs is shown as dotted line (right y-axis), and has been multiplied by the calving flux for each glacier (dashed line;

right y-axis).

of the seven glacier systems mentioned above was assigned a provenance sector (Prydz Bay sector: Lambert, Philippi and Denman glaciers; Wilkes Land sector: Adams, Totten, Thompson and Frost glaciers; Figure 5.1). From this, the proportion of icebergs that melt at each site sourced from either sector was calculated and multiplied by modern day calving fluxes (Gladstone et al. 2001), allowing for comparison with published and new IRD provenance data from core-top sediments at these sites (Brachfeld et al. 2007; Roy et al, 2007; this study).

Results are presented in Figure 5.8, where modelled and measured provenance data are expressed as ratios of Wilkes Land icebergs/IRD over Prydz Bay icebergs/IRD and show an excellent agreement for four of the five sites. This good agreement indicates that even though IRD loading is not taken into account, and wind field strengths were unchanged, we can be confident that the model is capturing the physical process that today govern the distribution of melting icebergs (and therefore their provenance at a given location) around the East Antarctic continent. The key factor controlling the proportion of melting icebergs supplied to any individual site is its location relative to source glaciers. For example, sites RC17-51 and ODP Site 1166 (Sites 1 and 3; Figure 5.8) are located down-current of the Lambert Glacier and as such the majority of icebergs that melt at these sites are sourced from this glacier system. Conversely, icebergs that melt over ELT47-17, ELT47-07 and ODP Site 1165 are mainly supplied from the Denman and Philippi Glaciers. One interesting observation when comparing modelled and observational data in the Prydz Bay area is that even though the amount of Wilkes Land-derived icebergs melting over ODP Site 1165 is low for the modern day, it is still higher than in any of the other sites (Figure 5.8). This observation can be explained by the more proximal locations of sites ODP Site 1166, 07 and RC17-51 to the continent, and the comparatively more distal setting of Site ELT47-14. The iceberg trajectory model hence reveals a zone of maximum westward transport, which is matched by the highest abundance of icebergs transported today (Antarctic Iceberg Tracking Database [1978–2012]; available at http://www.scp.byu.edu/data/iceberg/

database1.html). Of all sites shown in Figure 5.8, ODP Site 1165 is located the closest to this region of maximum iceberg transport (indicated by the thicker arrows in Figure 5.8), implying it is ideally placed to study far-field IRD supply back in time.

The only site to show any significant disparity between modelled iceberg melting patterns and observed IRD provenance patterns is ODP Site 1166. The unrealistically high Wilkes Land iceberg contribution obtained from the model run may be an artefact of small

scale ocean circulation features that occur at the front of the modelled Amery Ice Shelf and are not well captured within the employed global circulation model.

In summary, the excellent agreement between the overall results in iceberg melting and IRD provenance suggest that the model successfully replicates the most important features that control modern iceberg trajectories and influence the ratio of Wilkes Land to Prydz Bay IRD delivered to ODP Site 1165, i.e. the location and direction of the coastal current.

5.8.2 Elevated sea surface temperatures and iceberg melting patterns

To test the effect of increased SSTs on modelled iceberg melting patterns, sensitivity experiments were performed with the same pre-industrial ocean currents and wind patterns as described above, but with uniformly increased SSTs. These results are then compared to observed Pliocene IRD provenance in order to monitor Pliocene iceberg melting patterns at ODP Site 1165 under warmer ocean temperatures. In multiple runs, SSTs were increased in 1°C increments up to 8°C above pre-industrial values (Figure 5.9a). Overall, the pattern of icebergs derived from all glaciers to melt above ODP Site 1165 (apart from the Lambert Glacier, see above) shows the same general trend with increasing temperature: an initial increase with increasing temperatures, followed by a subsequent decrease with further increasing temperatures (Figure 5.9a). This result suggests that as SSTs increase, iceberg melt rates over the site increase, before a temperature threshold is reached and a significant fraction of icebergs melt before they can reach ODP Site 1165. It would be expected that the distance of each glacier from ODP Site 1165 would therefore play an important role in how many icebergs could survive before melting over the site, and that if this was the case, icebergs from more distal sites would reach their peak melt rate at a lower temperature than icebergs produced from more proximal glaciers. This pattern is exactly what is observed in the modelled melting patterns of icebergs produced from each glacier, apart from the Thompson Glacier, which produces the highest melt percentages at ODP Site 1165 for similar to present day temperatures of all Wilkes Land glaciers (Figure 5.9a).

Converting the modelled iceberg melting rates for all combined Prydz Bay glaciers and Wilkes Land glaciers with increasing SSTs to the ratio of Wilkes Land icebergs over Prydz Bay icebergs (i.e. comparable to the IRD provenance data in Figure 5.5), shows there is a clear trend, whereby more Wilkes Land icebergs are observed to melt over ODP Site 1165 at colder temperatures between 0°C and 2°C than at warmer temperatures (3 to 8°C)

(Figure 5.9c). These results can explain to a large extent why the Late Pliocene decline in SST around ~3.3 Ma (Table 5.5) was accompanied by a shift in the observed IRD provenance pattern, to a more distally sourced Wilkes Land-bearing assemblage.

5.9 Evidence for Increased Iceberg Flux from EAIS Destabilisation Events During