MicroCT+bubble+measurements

=**First porosity estimates from microCT measurements**=

The porosity that we measured with the MicroCT is about 5% less than what we got through gravimetric density measurements in McMurdo. The 5% don't make such a big difference in samples with higher porosity like firn, but they make a big difference (100%) in porosity of samples which have a vey low porosity to start with. Martin and I think that this is because we underestimated the density in McMurdo; the samples that we cut in McMurdo always have some surface roughness and this surface roughness will result in apparently more pore space.

The samples that we measured in McMurdo were not perfect cylinders. We measured the diameter and hight all around the cylinder and took the average to try to account for that, but the Volume of the cylinder is probably still smaller than what we measured with the caliper. This is likely to be cause of the roughness of the sample from cutting on the bandsaw. (We likely always measured the top of the roughness elements, [|CaliperMeasurementErrors.pdf]).

The following 2 figures show examples of the size distribution of pores in surface samples (sample # correspond to core numbers):



Figure below shows connectivity of the ice structure in the sample, indicating that the pores are still connected:



=MicroCT results from Allan Hills Transects=

The evaluation of the micro-ct results from the Snowball Project Allan Hills snow/firn/ice is summarized in [|MicroCT_Results_WithCrackEvaluation.xls] For all cores, we have measured the top 0-2 cm. For a firn core (26a) and for an ice core (38a), we made measurements with depth to see if they vary, which they don't seem to, and we can assume that the cores are more or less uniform with depth, at least when it comes to density. The SSA varies a bit more, but I am not sure how large an influence on the albedo these variations might have.

The SSA in the spreadsheet is bubble surface area either per weight or per volume of ice. r_bubble (bubble radius), n_bubbles_tot (total number of bubble in sample) and bubble density (number of bubbles per mm^-3 of sample) are calculated using the total bubble surface and the total bubble volume (for calculations, see [|BubbleEffectiveGeometry.pdf]), and they have the same SSA as calculated from the micro-ct (I think they therefore can be used for effective bubble properties).

(r_eff snow and n_eff snow are the effective radius and effective bubble number density of snow grains (if we had measured snow), and they have no implications for the ice albedo. I included them just for 'fun':-).)

"Ave. bubble diameter from mct-model (mm)" (as well as the standard deviation and the median) in the spreadsheet is the average bubble size from the micro-ct model. The model fills the air pockets with spherical bubbles, so if air space is not a prefect bubble, the model will fill it with the largest possible sphere and then fill the space around it with smaller spheres (like in this figure ). This is therefore not a measure of how many air pockets there are and how big they are, but rather a measure of something like the "average thickness of of the air" in the sample. This is also how the bubble-size distribution is calculated. If the bubbles were perfect spheres, then this model would represent average bubble size.

=Ideas on influence of cracks=

The most likely reason for the cracks that are present in the Allan Hills ice the cracking through thermal expansion. The daily temperature gradients are probably insignificant, but the seasonal temperature gradients are around 30K (just a guess, but I assume that is not too far off). The penetration of the seasonal temperature signal is around 4 m and no significant cracking should be found below. We measured core 38a on a 10 cm resolution and there were cracks in most samples. We have also measured samples from Kurbatovs group and none of their samples (the shallowest is around 10 m) shows such cracks. Now, on equatorial snowball earth, there should be no seasonal signal, because it's on the equator. Abbot 2010 ([|Abbot2010b.pdf]) shows that even the daily temperature variations are very small, if the model resolution is high enough (~ 5 K with 60 layers and 30 K with 2 layers). But either way, the penetration of the daily cycle in ice is only around 17 cm. I also calculated the thermal stress for different temperature cycles and if the daily cycle is 5 K, the thermal stress is around 2 MPa, which is around the tensile strength of ice (depending which values I use), and it is possible that there would be no cracking even at the surface (top 20 cm).

Apart from the cracks, the seasonal temperature cycles in the Allan Hills could also cause bubble migration or at least the changes in bubble morphology to make them as irregular as we find them at the surface (compared to how they look at 10 m depth). Both the cracking and the changes in morphology lead to a decrease in density and a higher SSA. This implies that the ice we find in the Allan Hills might have a higher porosity and a higher SSA than equatorial blue ice during Snowball Earth would have.

To estimate the contribution of the cracks to SSA, we can evaluate the cracks separately from the rest of the sample and see what percentage of SSA they make. The only idea I have to estimate the contribution of the changes in bubble morphology through temperature gradient migration is to compare the SSA of a non-cracked sample to the SSA of Kurbatovs shallowest sample. If we do this and can take the effect of these two processes that are driven by seasonal/daily temperature cycles out, we can then put a lower boundary for SSA and a lower possible boundary for the albedo on the equatorial ice on Snowball Earth.

**Bubble Size Distribution**
I plotted the Bubble Size Distribution for the MicroCTed samples in Figure [|AH_BubbleSizeDistribution_DifferentSnowFirnIceTypes.pdf]. There are 6 different figures, showing the transect and the different snow/firn and ice types, as well as the bubble-size-distribution with depth for ice core 38 and for firn core 26. There is also a comparison of the blue ice types including core 38, a sample from Shimmering Icefield, a sample from Byrd glacier surface and a sample from 10 m depth in the Allan Hills (from Kurbatovs core). The big difference in the Kurbatov Core bubble size distribution and all other surface ice samples is likely to be because of the seasonal temperature cycles (see above).

The bubble size distribution was calculated with the micro-ct model, that I described above (I believe that is also called "maximal ball algorithm").

=Deep core microCT profile=

11 samples will be analyzed from I-349 deep core (called Site 27). Below is a link to the file containing sample numbers, measurement numbers, sample names, and depths. Samples were cut during the week of 1/8/11 to 20mm diameter and run on microCT 80.