We present an analysis of the distribution of bluefin tuna Thunnus thynnus schools spotted during aerial surveys in the Gulf of Lions, in relation to oceanographic features. Bio-optical and thermal properties of the sea surface derived from high-resolution sensors (AVHRR and SeaWiFS) were studied on a daily basis, and an edge-detection technique was applied to detect frontal zones. Geostatistics and point-process analyses were used to evaluate the role of the environment in structuring the spatial pattern of bluefin tuna (BFT). The distribution of schools spotted was strongly non-stationary both in space and time; this is believed to be an effect of the survey design (transect sampling) and the influence of transient oceanographic structures (surface fronts and eddies). The empirical variograms indicated a spatial range of the BFT schools at around 40 km, with substantial daily variability. Ripley's K statistic, as well as autocorrelation plots, revealed that the fish schools were clustered over a wider range of scales (from 10 to 80 km), indicating more spatial structure than would be expected from a random process. Finally, BFT school distributions appeared well determined by the oceanic features, except at very small scales (<10 km), where over-aggregation occurred, and at the largest scales of our study (>40 km), where over-spreading was detected. Dynamical ecological processes, such as foraging, are likely to induce this complex spatial pattern. Possible reasons for the association of tuna with fronts are presented.