Abstract. The objective of this work was to evaluate the relative trap effi ciency of the three modifi ed traps baited with 1-octen-3-ol as attractant, Box, Nzi and Canopy, for horse fl y (Tabanidae) survey. Nine traps (three traps per trap type) were tested in 3 × 3 Latin square designs during summer 2011 in riparian oak-ash forests in the Danube fl oodplain of Croatia. Overall, the traps caught 1,986 specimens of 11 horse fl ies species during the study period. The most abundant species caught was Tabanus bromius (58%), followed by Tabanus tergestinus (21%), Tabanus maculicornis (16%), Tabanus sudeticus (2%), Atylotus loewianus (1%) and Tabanus autumnalis (1%). Both the Box traps and the Nzi traps had signifi cantly greater catch effi ciencies than the Canopy traps (β box = 2.47, X 2 = 65.48, df = 1, P < 0.001 and β nzi = 1.54, X 2 = 25.12, df = 1, P < 0.001, respectively). Nzi traps had a catch rate 4.65 (95% CI: 2.55, 8.48) times greater than Canopy traps and Box traps had a catch rate 11.77 (95% CI: 6.48, 21.39) times greater than Canopy traps. Based on Shannon-Wiener diversity indices Nzi traps were better suited for diversity survey and had higher (H' = 1.33) indices than Box (H' = 1.08) or Canopy traps (H' = 1.00).