Faults which can compartmentalize a reservoir and which are too small to be seen on seismic can be predicted using a self-similar (or fractal) model of faulting at various scales. A computerized method has been developed for statistical analysis of a fault pattern interpreted from three-dimensional (3-D) seismic in order to fractally derive the distributions of fault size, frequency, and spatial variations and to verify their invariance within the seismic-resolution range. This analysis can be used for quality control of seismic interpretation and as input for extrapolation toward smaller scales using probabilistic techniques. The application of this program to Pelican field, United Kingdom North Sea, validated the interpretation of the complex fault pattern of that field but suggested that only limited extrapolations could be confidently made. Probabilistic maps showing distribution of faults down to 3 m (10 ft) of throw and not resolved by 3-D seismic were made. These maps subsequently were used as a qualitative aid in full-field reservoir simulations to model differently zones of potential dense faulting from zones of reduced faulting. This was done because nearby field behavior had shown that faulting is a significant factor in reservoir behavior and ultimately field deliverability. Field development planning, therefore, could be optimized.