Carbon stable isotopes can be used to trace the sources of energy supporting food chains and to estimate the contribution of different sources to a consumer’s diet. However, the δ13C signature of a consumer is not sufficient to infer source without an appropriate isotopic baseline, because there is no way to determine if differences in consumer δ13C reflect source changes or baseline variation. Describing isotopic baselines is a considerable challenge when applying stable isotope techniques at large spatial scales and/or to interconnected food chains in open marine environments. One approach is to use filter feeding consumers to integrate the high frequency and small-scale variation in the isotopic signature of phytoplankton and provide a surrogate baseline, but it can be difficult to sample a single consumer species at large spatial scales owing to rarity and/or discontinuous distribution. Here, we use the isotopic signature of a widely distributed filter-feeder (the queen scallop Aequipecten opercularis) in the northeastern Atlantic to develop a model linking base δ13C to environmental variables. Remarkably, a single variable model based on bottom temperature has good predictive power and predicts scallop δ13C with mean error of only 0.6 ‰ (3%). When the model was used to predict an isotopic baseline in parts of the overall study region where scallop were not consistently sampled, the model accounted for 76% and 79% of the large-scale spatial variability (101–104 km) of the δ13C of two fish species (dab Limanda limanda and whiting Merlangus merlangius) and 44% of the δ13C variability in a mixed fish community. The results show that source studies would be significantly biased if a single baseline were applied to food webs at larger scales. Further, when baseline δ13C cannot be directly measured, a calculated baseline value can eliminate a large proportion of the unexplained variation in δ13C at higher trophic levels.