Fish scales of Salmo trutta exhibited regional patterning in Sr and Mn concentrations in major third-order tributaries in a small upland catchment (the Dee catchment in Wales, United Kingdom; drainage area ∼1800 km2) that appeared to reflect regional variability in catchment geology and stream water chemistry. When baseline signatures in scale element concentrations were established at a number of sites (n = 12) among the upper, middle, and lower regions of the Dee catchment, 73% of fish were classified to their region of capture based on Sr and Mn in scales. However, the regional patterning in scale chemistry was degraded when high-resolution catchment-wide variability in scale element concentrations was predicted using element concentrations in stream water as a proxy for Sr and Mn in scales at 792 sites in first-, second-, and third-order tributaries in the catchment (mean distance between neighbouring sites = 738 m, range = 41–3634 m). This analysis indicated that many locations throughout the catchment could be potential source locations and demonstrates that the initial classification accuracy was artificially inflated. We have illustrated limitations of the site-based study design commonly employed in biogeochemical tagging studies. Future studies should take account of potential variation in baseline signatures at fine geographical scales when determining the accuracy of stock discrimination using biogeochemical tags.