Towards understanding the signal formation in polar snow, firn and ice using X-ray computed tomography

Polar ice cores act as a unique archive of the Earth's climate system. However, due to logistic constraints, the representativity of these ice-core records cannot be estimated directly. One possible remedy is to analyze the spatial variability in polar snow and firn and combine the results with an improved understanding of the formation of paleoclimatic ice-core signals and their evolution with depth. Here, X-ray computed tomography is applied as a non-destructive method that yields information on stratigraphy and microstructure in polar snow and firn. The results are used to contribute to both subtopics of this indirect approach for estimating representativity. New methods for sampling the snowpack as well as the detection and alignment of coherent signals in spatially-distributed datasets are presented. They are applied to analyze spatial variability in the snowpack both on the local (trench studies in Greenland and East Antarctica, distances up to 100 m) and the regional scale (450 km traverse through North Greenland). The matching algorithm is validated using randomly generated profiles with the same statistical properties as the original data. Snow and firn density as markers of stratigraphy are determined by two-dimensional radioscopic imaging, the water-isotopic d18O signal is used for age dating. The results show that regionally a significant share of the stratigraphic density signal persists over hundreds of kilometers. Locally, there is a strong directional influence of the wind with a much larger homogeneity of the snowpack along the main wind direction. As density is an important input parameter for remote sensing and surface-mass-balance estimates, representative profiles or mean values of snow and firn density are required. Such a profile is provided for the upper two meters of the North Greenland snowpack. On the local scale, the estimation of representative densities for certain areas of interest (such as the footprint of an altimeter) is complicated by the directional dependence of the stratigraphic variability. As the density layering is significantly impacted by melting of the snow surface, melt features dating to the warm Greenlandic summer of 2012 are analyzed in detail. A large heterogeneity of these features is quantified, which does not only affect remote measurements (where ice layers act as reflectors for electromagnetic waves) but also strongly influences the ability to interpret single-core melt records. Methodological advances in the three-dimensional computed tomography of polar firn allow the creation of a first extensive dataset of direct firn-microstructure measurements. Three ice cores that represent different extremes of the temperature and accumulation ranges are analyzed throughout the lock-in zone, the depth range where pores are sealed from the atmosphere. The fundamental lock-in process is a determining factor for the gas-age-ice-age difference, which can be on the order of several 1,000 years. Thus, accurate estimates of this value are of particular importance for the interpretation of phase relationships between ice and gas records. The dataset is used to show that the critical porosity of pore enclosure is a climate-independent constant, a finding that is corroborated by percolation theory. Incorporation of this result significantly influences the dating of trace-gas records, reducing mismatches with other climate proxies by up to more than 1,000 years. Furthermore, it is demonstrated why previous measurements yielded misleading results