Estimation de la précision des campagnes acoustiques au Sénégal par la méthode géostatistique transitive à une dimension

Abstract

This study presents an application of the geostatistical one-dimensional (ID) transitive method for estimating the precision of acoustic pelagic abundance estimates in Senegal. In applying the method, the major problem encountered was how to take into account the temporal variability in the computation of the estimation variance. In acoustic surveys, echointegration is performed continuously along the ship's sailing track. For a survey design made of parallel transects, each transect can thus be considered as a sarnpling unit. By summing the dcnsity values along the transects to ohtain biomass values per transect, the 2D estimation is sirnply rcduced ta a ID estimation procedure. If the transects are equidistant, the gcostatistical transitive mcthod applied in ID on the biomass values per transect gives a correct and simple solution to the estimation of the survey prccision. The method has bccn applied to the Senegalcse acoustic surveys which were performed with the same sarnpling design made of regularly spaced parallel transects oriented East-West. The densities were summed along the transects. The latitudinal variation of the biomass per transect was presented as a ID profile. Biomass, spatial distribution and its structure (autocovariance) showed great differences between surveys. Profiles were grouped in 4 categories which presented different structural characteristics. In each category, the average structure was estimated and modelled and the average relative estimation error was estimated. The relative estimation error varied with the differences in structure between categories but also with mode1 hypothesis about the structure at small distance. Two hypothesis were considered, continuity of the distribution at small distance (absence of nugget effect), or discontinuity (presence of nugget effect). The maximum relative error estimated was 31%. The major problem in modelling the structure of the ID profile was to fit or not to fit a nugget effect. The interpretation of such nugget could be temporal variability at the scale of the transect.