We use polarimetric radar sounding to investigate variation in ice crystal orientation fabric within the near-surface (top 40-300 m) of Rutford Ice Stream, West Antarctica. To assess the influence of the fabric on ice flow, we use an analytical model to derive anisotropic enhancements of the flow law from the fabric measurements. In the shallowest ice (40-100 m) the azimuthal fabric orientation is consistent with flow-induced development and correlates with the surface strain field. Notably, toward the ice-stream margins, both the horizontal compression angle and fabric orientation tend toward 45 degrees relative to ice flow. This result is consistent with theoretical predictions of flow-induced fabric under simple shear, but to our knowledge has never been observed. The fabric orientation in deeper ice (100-300 m) is significantly misaligned with shallower ice in some locations, and therefore inconsistent with the local surface strain field. This result represents a new challenge for ice flow models which typically infer basal properties from the surface conditions assuming simplified vertical variation of ice flow. Our technique retrieves azimuthal variations in fabric but is insensitive to vertical variation, and we therefore constrain the fabric and rheology within two end-members: a vertical girdle or a horizontal pole. Our hypotheses are that fabric near the center of the ice-stream tends to a vertical girdle that enhances horizontal compression, and near the ice-stream margins tends to a horizontal pole that enhances lateral shear. ApRES radar data were collected as part of the BEAMISH Project (NERC AFI award numbers NE/G014159/1 and NE/G013187/1). Tom Jordan would like to acknowledge support from EU Horizon 2020 grant 747336-BRISRES-H2020-MSCA-IF-2016. ***** PLEASE BE ADVISED THIS DATA SET HAS BEEN RETRACTED ***** This data set had incorrect coordinates for one of the sites. In addition, some files were incorrectly labelled as belonging to one of the sites A new data set (see 'Related Data Set Metadata' link below) addresses these issues and also includes significant additional data, as well as updated metadata and additional authors. Hence it is a wholly new data set, rather than an updated version of this data set. Please use this new data set instead.

These two files (.csv) provide independent methods of quantifying subglacial roughness in Greenland, both calculated from radio-echo sounding (or ice penetrating radar) data collected by the Operation Ice Bridge programme using CReSIS instrumentation. They are an output of the Basal Properties of Greenland (BPOG) project (http://bpog.blogs.ilrt.org/), with funding from NERC grant NE/M000869/1. Roughness here, and in the wider literature, is defined as the variation in bed elevation (in the vertical) at the ice-bed interface, over a given length-scale. These two metrics calculate/quantify this variation in different ways: one shows topographic-scale roughness, calculated from the variation in along-track topography (bed elevation measurements derived from the radar pulse); and the other shows scattering-derived roughness, calculated from quantifying characteristics of each bed-echo (the return from the radar pulse at the ice-bed interface).