Why the number of hook per basket (HPB) is not a good proxy indicator or the maximum fishing depth in drifting longline fisheries ?

The effects of targeting in longline fisheries have major impacts on CPUE (catch per unit effort). To target large pelagic fishes in longline fisheries, fishermen deployed several tactics according to the target species. For a given species, the tactic depends on both the season and the fishing ground. One of major aims of the fishing tactic is to set the longline for optimizing the overlap between the distribution of hook depths and the supposed preferential vertical habitat of the target species. Thus, the maximum fishing depth (MFD) of the longline is recognized as a key parameter for estimating the distribution of hook depths and then to standardize fishing effort. In this context, the number of hooks per basket (HPB) is commonly used as a proxy indicator of MFD (this information has been collected for the Japanese longline fishery since 1975). Unfortunately HPB is not a good proxy indicator of MFD. In the first part of this work we illustrate why our previous affirmation is consistent from a theoretical point of view. Second, we analyse depth series recorded with time depth recorders (TRDs) at the deepest point of the mainline (data collected during instrumented longline experiments carried out in the French Polynesia EEZ, central South Pacific). We observe variations of depths recorded during the soak time and we compare the calculated mean depth of depth series with the theoretical MFD (catenary formula). The mean depth appears as the better proxy indicator of the MFD which can be used to estimate the distribution of hook depths. Finally, we discuss relevant data which must be collected by both observers and fishermen in order to improve analysis of the fishing effort in longline fisheries. These data are crucial to standardize appropriately longline CPUE and to correct for potential bias. Statistical models (even the most sophisticated, such as habitat based model, GLM, GAM, GLMM or neural networks) could not supplant for the lack of such input information