Recrystallization regimes in an ice sheet - Towards a microstructure-based law of ice
Zusammenfassung der Projektergebnisse
A digital image processing approach has been developed to extract microstructural information from large sets of sublimation groove images acquired along deep polar ice cores. The method has been applied to image data from the EDML (East Antarctica) and NEEM (Greenland) ice cores. A particular characteristic of this analysis technique is the high resolution of the mapped grain boundary networks. This represents a significant improvement in comparison to previously applied methods like c-axes orientation measurements. The automatic extraction of grains and grain boundaries enables a fast parameterization of grain boundary networks, which in turn allows for a high vertical resolution along the core. With the novel method mean values of grain size, grain shape and the shape of grain boundaries can be derived in a consistent way from a large amount of data. Due to the higher vertical resolution of grain geometry parameters presented in this study, it becomes clear that previous assumptions about strength and variability of internal stresses have to be questioned. The results support recent studies which reported that the dominating recrystallization processes are not only dependent on depth. In this context, the development of a preferred orientation of c-axes plays an important role. For all sections of the NEEM ice core for which sublimation grooves have been mapped, the orientation of c-axes has been acquired subse- quently. The analysis of the c-axes orientation by means of the here presented image processing techniques is subject of a Diploma thesis. The availability of highly resolved grain boundaries and the orientation of c-axes for the same set of ice core sections provides the possibility to compare the results from these two very different methods and to estimate possible errors. An automatic matching algorithm has been developed to estimate errors in the parameterization of grain boundary networks, which arise as missing or falsely detected grain boundaries during acquisition and processing of sublimation groove images. A long-term perspective is to systematically assign c-axis misorientations to reliably extracted, highly resolved grain boundaries from sublimation grooves. For six automatically matched NEEM image pairs the area fraction of grain crosssections reliably assigned to c-axes orientation lies in the range of 50-80%. The presented matching algorithm has been found to be an effective tool for error estimation and for identifying problems in the assignment of grains and grain boundaries. During this study several ideas to improve the image acquisition method have been developed. Unpredictable sublimation behavior can be reduced if grooves on the polished ice surface evolve in a chamber of defined climatic conditions, which is technically feasible even at drill sites. Manual adjustment of the focus is no longer necessary if ice core sections are cut to uniform thickness. Subsequent to mapping of sublimation grooves, thick sections of the NEEM ice core have been further processed to thin sections. It is assumed that sublimation grooves mapped on the surface of thin sections exhibit less deviations in comparison to the measurement of c-axis orientation, which have to be performed on thin sections by a fabric analyzer in transmitting light. As the latter is capable to map sublimation grooves at low resolution, the application of both acquisition methods on thin sections can reduce inaccuracies associated with the present matching algorithm. Every processing step on the sublimation groove images can be verified by / compared with the extracted grain boundaries from measurements of c-axes ori- entation. Occasionally, the algorithm to remove vertical scratches (caused during polishing of the surface) label grain boundaries as artifacts. Based on the matched image pairs, the influence of this algorithm on the agreement of extracted grain boundaries has been quantified. Generally, it has become possible to reconsider the choice of parameters and thresholds used throughout the image processing approach in a systematic way. In addition to grains and grain boundaries, also sub-grain boundaries have been extracted from sublimation groove images by the developed image processing approach. However, this part of the processing is in an early stage of development. A comparison between automatic and previously performed manual extraction shows that typically only half of the sub-grain boundaries is detected. This discrepancy has been traced back to the inadequacy of the present method to separately detect sub-grain boundaries in close vicinity. As solving this problem by means of anisotropic diffusion is associated with a drastic increase in computing effort, it seems more appropriate to first examine whether improved image acquisition is sufficient. As the sublimation grooves at the site of sub-grain boundaries are represented by gray values similar to those inside grains, an increased contrast could enable their automatic extraction. Owing to the fine resolution in the position of grain boundaries, much smaller grains can be extracted from sublimation groove images in comparison to c-axes measurements. This raises the question how many of the small grains should be considered for the calculation of the mean grain size. A lower cut-off has been defined based on the area fraction of considered grain cross-sections to limit the influence of small-grain artifacts. A comparison of grain growth rates derived from the EDML and NEEM data sets by means of this cut-off to previously derived growth rates suggests that more small grains than previously are considered. This allows measuring the development of a second mode of smaller grains which has previously been identified, but not measured. Comparison of grain geometry data from the NEEM ice core with results from continuous flow analysis revealed a correlation between the occurrence of small grains and the dust content at 320 m depth. 10 meters below, a thin band of larger grains (increased by a factor of 10) appears to correlate with a high concentration of ammonium. These examples highlight that one mean values in grain size for an entire section or even a wider range do not reflect the mechanism- related variability on mm and cm scale adequately. The evolution of grain shape with depth has been analyzed. Changes in trend can partially be attributed to the presence of effective negative pressures caused by air bubbles and clathrate hydrates. Significant differences between the profiles of the NEEM and EDML ice cores are observed. Apart from methodical differences, deviations can be explained by the difference in age scales between both ice cores, which suggest that recovery is more effective along the EDML ice core. The obtained densities of sub-grain boundaries (derived from sublimation grooves) confirms this conclusion which is in good agreement with a previous comparison of the EDML ice core to experimentally deformed ice. For the NEEM ice core, high sub-grain boundary densities coincide with stratigraphic layers visible in radar profiles. Those layers are characterized by a high concentration of impurities, which confirms that a sound understanding of the coupling between different aspects of multi-scale variability is required to develop an improved micro-dynamical basis for mean-field approaches in ice sheet modeling. The presented method enables an objective quantification of consistency and deviations between microstructural observations in nature, deformation experiments and simulations.
Projektbezogene Publikationen (Auswahl)
-
Eemian interglacial reconstructed from a Greenland folded ice core,
2013, Nature 493, 489-494
NEEM community members
-
Extraction and parameterization of grain boundary networks in glacier ice, using a dedicated method of automatic image analysis,
2013, Journal of Microscopy 250, 130-141
T. Binder, C.S. Garbe, D. Wagenbach, J. Freitag, and S. Kipfstuhl
-
Microstructure through an ice sheet,
2013, Materials Science Forum 753, Recrystallization & Grain Growth V, 481-484
T. Binder, I. Weikusat, J. Freitag, C.S. Garbe, D. Wagenbach, and S. Kipfstuhl
