This paper introduces a refined trend surface analysis (TSA) for detecting faults with small, metre-scale offsets (5–20 m). Conventional TSA uses a global polynomial method to model the trend; the power of the polynomial (e.g., first, second, or third order) is the only parameter for input. The refined approach is different in that it uses local-fit techniques to generate the trend. The refined approach provides greater flexibility for inputting geological knowledge in the trend surface modelling process. This results in a trend surface with the maximum amount of unwanted information, which is removed from the residual surface after removal of the trend from the data, leaving features of interest optimally highlighted in the residuals. Such a trend is referred to as a geologist-controlled trend to differentiate it from the trend surface modelled by conventional TSA, which is primarily a computer-controlled global polynomial surface. The refined approach goes one step beyond conventional TSA by incorporating advanced geostatistics for modelling the trend, interpolating the resultant residuals, and then extracting formation-top offset patterns from the residual surface using spatial analysis. Modelling the geologist-controlled trend in the refined approach results in higher resolution in detecting formation-top offsets and higher accuracy in digitizing fault locations, compared with various techniques that have been traditionally used in subsurface structure mapping of the West Canada Sedimentary Basin (WCSB) (e.g., structural and isopach contour mapping, cross-section construction, and seismic and aeromagnetic data interpretation). The refined approach is demonstrated using the Basal Fish Scale Zone in the Peace River Arch region as an example.