Characterizing fractured reservoirs is challenging, and reliable inputs for fracture modelling are difficult to acquire and validate. We demonstrate that the inherent sparseness and directional bias of borehole data make it particularly problematic to derive meaningful estimates of fracture scaling, connectivity, and clustering from the sub-surface, yet each of these parameters has a major influence on the results of flow modelling. Core, borehole images and well-logs can improve understanding of the mechanical stratigraphy of a reservoir, and provide critical information to test the suitability of an analogue; the rich datasets derived from outcrop can then be used to improve the correlation between mechanical stratigraphy and fracture stratigraphy in the sub-surface.Modern geospatial methods including lidar and photogrammetry, both ground-based and from UAV (drone), have made it possible to acquire a 3D surface image of large areas of fractured outcrop, cheaply, quickly and safely. The resultant data are processed very rapidly to produce an oriented, scaled and georeferenced virtual copy of the outcrop that can be picked semi-automatically to derive copies of the fracture network containing tens or hundreds of thousands of individual fractures that collectively provide a basis for robust estimation of fracture network properties. This presentation will emphasize the importance of not relying on sub-surface data alone when characterizing fractured reservoirs, and will demonstrate the sensitivity of key fracture parameters on predicted reservoir performance. We illustrate this with large fracture models from outcrop analogues of fractured carbonate reservoirs from the Kurdistan region of the Iraqi Zagros, and fractured shales from the UK.