For effective flow-simulation models, it may be important to estimate permeability accurately over several scales of geological heterogeneity. Critical to the data analysis and permeability prediction are the volume of investigation and sampling interval of each petrophysical tool and how each relates to these geological scales. We examine these issues in the context of the As Sarah Field, Sirte Basin, Libya. A geological study of this braided fluvial reservoir has revealed heterogeneity at a series of scales. This geological hierarchy in turn possessed a corresponding hierarchy of permeability variation. The link between the geology and permeability was found to be very important in understanding well logs and core data and subsequent permeability upscaling. We found that the small scale (cm) permeability variability was better predicted using a flushed-zone resistivity, R(xo), tool, rather than a wireline porosity measurement. The perm-resistivity correlation was strongest when the probe permeabilities were averaged to best match the ''window size'' of the wireline R(xo). This behavior was explained by the geological variation present at this scale. For the larger scale geological heterogeneity, the production flowmeter highlighted discrepancies between flow data and averaged permeability. This yielded a layered sedimentological model interpretation and a change in averaging for permeability prediction at the bedset scale (ms-10 x ms).