Most of the recent framework directives and environmental policies argue for the development and the use of indicators - notably trophodynamic indicators - that should be able to follow ecosystems' evolution in space and time, particularly under anthropogenic perturbations. In the last decades, the use of stable carbon and nitrogen isotopes ratios has increased exponentially, particularly in studies of marine ecosystems' trophic structure and functioning. This method is principally based on the assumption that the isotopic composition of a consumer directly reflects that of its food. Nevertheless, few studies have attempted to define the limits of this tool before using it and drawing ecological conclusions from isotopic analysis. This study aimed to assess the importance of considering spatio-temporal variations in isotopic signatures of consumers when using δ13C and especially δ15N values in open ecosystems with complex food webs, using the Bay of Biscay (north-east Atlantic) as a case study. To this end, more than 140 species from this marine ecosystem were analysed for the isotopic signatures in their muscle tissue. They were sampled from coastal to oceanic and deep-sea areas and at different latitudes, to evaluate spatial variations of isotopic signatures. Selected species were also sampled over several years and in two seasons to account for inter-annual and seasonal variations. In the Bay of Biscay temperate ecosystem, which is subject to both coastal and oceanic influences (two main river inputs and upwelling areas), δ13C and δ15N values significantly decreased from inshore to offshore species and to a lesser extent from benthic to pelagic organisms. River discharges appeared to be the first factor influencing δ13C and δ15N values in consumers. From the important spatial variations detected in δ15N values in particular, we suggest that in such contrasted ecosystem, nitrogen isotopic ratios may also be revisited as an indicator of the feeding area. Moreover, we demonstrate that several baselines should be used when calculating trophic levels from δ15N values. From the temporal variations detected, we recommend concentrating on a short time scale for the sampling of most organisms.