Are bubbles in ice at a lower or higher pressure than atmospheric pressure? When water phase-transitions into ice it expands. The water usually contains dissolved air. Freezing forces the air out of the solution into bubbles.
Are these bubbles at a lower or higher pressure than atmospheric pressure? 
I can see arguments both ways: when a washer expands, both the outer and inner radius increase. This would imply bubbles are at a lower pressure, but perhaps the ice expands differently and squeezes the bubble and it is at a higher pressure. However, if they were squeezed I would expect them to try to expand, and propagate along weak fraction lines in the ice, and I do not see this - bubbles are usually round, suggesting low pressure environments.
Edit: Two papers on the subject, but neither sheds direct light on the answer:


*

*"Bubbles and bubble pressures in Antarctic glacier ice" (A. J. Gow, J. Glaciol. 7 no. 50 (1968), pp. 167-182) shows that the bubbles shrink with increasing pressure. These are therefore presumably high-pressure bubbles. However, the internal bubble pressure is due to ice compressing as more ice forms on top, not just the freezing process.

*"Air Bubbles in Ice" (A. E. Carte, Proc. Phys. Soc. 77 no. 3 (1961), p. 757) talks about bubble formation, but not internal bubble pressure.
 A: Fluid inclusion analysis techniques are used by geologists to gather information on the pressure, volume, and temperature conditions during the crystallization of the mineral (here ice) containing the inclusion.
There are three assumptions that usually made in dealing with fluid inclusions:
•   The composition of the trapped fluid has not changed since inclusion formation
•   The density of the trapped fluid has not changed since inclusion formation
•   The volume of the inclusion has not changed since inclusion formation
Natural fluid inclusions will contain multiple chemical components (impurities) and may contain multiple phases (gas, liquids, precipitated crystals).
The analysis of fluid inclusions (say to determine the pressure of formation) therefore involves measuring the composition and density and then applying the appropriate equation of state calculations to calculate unknowns.
Fluid inclusion measurements and analysis are not simple to perform.  I would not think you can say whether the pressure is higher or lower than the formation pressure without doing a complete analysis.
The analysis of air bubbles in ice are important for the study of the past composition of the atmosphere and climate change.    
A: Since the air in the bubbles was below the surface of the water before the freezing, I would expect it to be above atmospheric pressure, albeit maybe just a little, depending on the hydraulic head.
A: This is how I imagine it would happen, As Ice begins to freeze, outer surface first followed by the inner surface, the concentration of the air in the water would increase.
As more and more of the ice begins to freeze, the water would eventually  be saturated with air, and hence be forced to from a bubble. The final size of the bubble is same as the cavity that is created when water freezes.
If you have a sphere full of water(with dissolved air), and it is frozen, outer to inner, We know the size of the cavity inside the ice. We also know the the total amount of air water can hold at 0 degrees Celsius. 
http://www.engineeringtoolbox.com/air-solubility-water-d_639.html, 
The numbers can be calculated more correctly, but just to get an idea, the website indicates that the solubility of Oxygen in air is 14mg/l at zero degrees.
14 mg at STP = (14/1000g)/(16 g)*22.4 = .0196 litre 
the empty space created when water freezes is ~8.3 pc(http://en.wikipedia.org/wiki/Ice), ~.08 litres of space per liter of water frozen.
Therefore one can conclude that the pressure in the air bubble is lower. The numbers will change if you add the total air dissolved in water(the conclusion will still probably hold. This calculation is for an atmosphere of oxygen at 1 atm)
Ref http://academic.keystone.edu/jskinner/Limnology/Water_Chemistry_LectureNotes.htm
A: I love this question so I want to add my opinion here. 
I would think that the column of atmosphere above the ice would contribute more pressure than the expansion pressure on an air bubble in an ice cube. But I have no data on this.
As we know, the decrease in temperature results in some moles of air coming out of solution as gas. This would decrease the volume of the solution--except that the gas is trapped and is still part of the total volume. Since the air occupies more volume when in this gas state, the pressure would be some magnitude.
At the same time, the crystal structure of water is forming an organized system, which increases the volume of the solution as it expands in all directions--noticeably outward. The ice typically freezes from the outside to the inside, which would provide some containment pressure on the gas bubbles. It would be interesting to compare the air bubble pressure at the core versus in a superficial layer.
As a test at sea level, I would set up high-definition/magnification cameras on air bubbles in ice, surrounding with a low viscosity, non-polar dye solution. I would then observe one bubble as its sphere is first broken upon melt. If there is a spurt of air from the bubble then we conclude the air bubble pressure is higher than atmospheric and if some of the color is observed to rush into the air bubble--which I would suspect--then we can conclude the bubble pressure is less than atmospheric.
Interestingly, the bubble pressure could turn out to be close enough to atmospheric pressure to not notice anything.
As an aside for further reading on a variant of this topic, look into the crystallization of the molecules in glass in a Prince Rupert Drop.
A: An answer for glacier ice is different than frozen water - the glacier ice has formed under different processes. Still, it is interesting to see the argument laid out by Pettit et al. (2015) that bubbles in glaciers are at higher pressure than atmospheric pressure.

As ice flows toward the terminus, overburden pressure decreases, the ice viscously relaxes, and pore pressure is reduced. However, once the difference in pressure between the pore space and the ice crystal lattice is less than about 10 bar, deviatoric stresses are insufficient to permit further pressure equalization due to the nonlinear rheology of ice [Gow, 1968]. Air-filled pores in melting glacier ice, therefore, will always be at a higher pressure than atmospheric pressure. Scholander and Nutt [1960] measured the pressure in the pores of west Greenland icebergs and found them typically not only between 5 and 10 bar but also up to 20 bar. 

From http://dx.doi.org/10.1002/2014GL062950
