The precise assessment of the catches by species is a major element in multi-species fisheries, such as the tropical
tuna purse seine fisheries. The species composition by set is reported in the logbook, but it has been evidence of
large bias mainly for the small individuals in the logbooks, which prevent the direct use of that source for catch
estimates. For the major tropical tuna purse seine fisheries operating in the Indian Ocean, the species composition
is estimated from sampling operations at landing and thought a statistical treatment to interpolate value for nonsampled
sets. This method, called the Tropical Tunas Treatment (T3), developed by IRD and IEO in the mid-1990s
has been criticized, specifically in the part on the species composition corrections. This document presents the
results of a new statistical approach to handle the different shortcomings pointed out using data collected from
the French fleet in the Indian ocean. Analyses specifically focus on the spatio-temporal dimension of the catches.
Furthermore, the use of more information from the logbook reports are investigated and discussed.