Outcrop analogues are an important tool for understanding the small scale architectures of structurally complex reservoirs. Recent developments in global positioning system (GPS), mobile computing and laser scanning technology allow efficient capture of digital data from outcrop analogues. The key advantage of digital acquisition over traditional paper-based mapping is the ability to record field data in their correct spatial (x,y,z) locations. Digital outcrop data are readily displayed and manipulated in standard reservoir or structural modelling packages, allowing the user to build three-dimensional (3-D) reservoir analogues by projecting geological planes into/out from the outcrop surface. It is also possible to combine digital field data with shallow subsurface ground penetrating radar (GPR) surveys to improve 3-D constraints. Nevertheless, a limitation of most modelling packages is that the interpreted 3-D geological architectures are represented by deterministic objects which mask the uncertainties inherent in the acquisition and modelling of the underlying data. The aim of this presentation is to investigate and, where possible, quantify the uncertainties associated with a digital model of sub-seismic scale, post-depositional normal faults (maximum throw < 15 cm) which cut a sandstone-shale sequence exposed on the foreshore at Lamberton, SE Scotland. We have mapped the footwall and hangingwall cut-offs of two prominent sandstone beds (< 50 cm thick) using Real Time Kinematic (RTK) GPS with a nominal horizontal precision of ~ 1 cm and vertical precision of ~ 2 cm. We have deliberately chosen to map faults with displacements that are close to the resolution of the RTK GPS, in order to highlight the impact of spatial uncertainty on interpreting digital models. Data were recorded every ~20 cm along the length of each fault, wherever possible starting/ending at fault tips. In addition, for each GPS position we also measured fault throws and fault plane orientation manually using a steel rule and compass-clinometer. Finally, we used the RTK GPS to create a digital elevation model (DEM) of the two faulted sandstone beds. All the point data were imported into a structural analysis package (Badleys TrapTesterTM) in order to create triangulated fault and horizon surfaces and to digitise fault polygons from the mapped horizon cut-offs. Fault orientations derived from the GPS-based digital outcrop model are broadly consistent with those measured manually in the field. However, where fault trace lengths are short and/or where either the hanging wall or footwall cut-offs are missing, the modelled fault dips are significantly lower than the measured dips; locally, the modelled fault surfaces may even appear to dip in the opposite direction to the natural faults. Analysis of the vertical and horizontal error bars (i.e. precision) of each GPS data point shows that heave and throw can only be resolved near the centres of faults, i.e. regions of a fault situated around the point of maximum displacement. Thus, fault polygon lengths are truncated by up to 50 % although in contrast to fault surfaces interpreted from seismic data the positions of all exposed fault tips are observed directly and are precisely located within the model. Throw populations derived from the digital dataset are therefore not exact equivalents of the throw populations derived from manual field measurements and should be corrected for truncation. As many previous studies have illustrated, digital survey methods have great potential for capturing and visualising quantitative 3-D outcrop analogue data. Here, we highlight the importance of calibrating the digital outcrop model against real field data. This enables us to understand and quantify the sources of uncertainty that arise during data capture and modelling. This calibration step is essential before digital outcrop data can be upscaled to provide quantitative analogues for the analysis of structurally complex reservoirs.