Past temperature reconstructions from polar ice sheets are commonly based on stable water isotope records in ice-cores. However, despite major efforts in the understanding of the ice-core signal formation, the temperature reconstructions of the last millennium in Antarctica remain highly uncertain. Here, using a 100 km scale representative surface water isotope dataset, we show that the spatial variability of local surface topography and accumulation rate anomalies influences the isotopic composition of the upper-meter snowpack. The magnitude of this non-temperature effect on water isotopes is similar to changes of the last millennium. We demonstrate that these spatial anomalies are advected into the deeper firn and ice column, and are translated into an artificial centennial to millennial scale variability in the isotope record. Additionally, we provide an estimation of areas where this effect is relevant for last millennium temperature reconstructions.

We analysed crystal-preferred orientation of c-axis and microstructure data from the Alpine ice core KCC at an unprecedented resolution and coverage of any Alpine ice core. We find that an anisotropic single-maximum fabric develops as early as 25 m depth in firn under vertical compression and strengthens under simple shear conditions towards the bedrock at 72 m depth. The analysis of continuously measured intervals with subsequent thin section samples from several depths of the ice core reveals a high spatial variability in the crystal orientation and crystal size on the 10 cm-scale as well as within a few centimeters. We quantify the variability and investigate the possible causes and links to other microstructural properties. Our findings support the hypothesis that the observed variability is a consequence of strain localisation on small spatial scales with influence on fabric and microstructure. From a methodological perspective, the results of this study lead us to challenge whether single thin sections from ice cores provide representative parameters for their depth to be used to infer the fabric development in a glacier on the large scale. Previously proposed uncertainty estimates for fabric and grain size parameters do not capture the observed variability. This might therefore demand a new scale-sensitive statistical approach.