One of the main limitations in acoustic stock assessment is fish identification. We propose here a new method for identifying several communities of tropical fish within a surveyed area. This method, called "Acoustic Populations", consists in splitting a study area in systematic regular strata such as geographical rectangles, and calculating for each stratum a set of acoustic parameters that can be easily obtained with conventional acoustic equipment, and have a discriminant power, such as target strength, mean density, confidence interval and dispersion indexes. It is based on the fact that the biological (specific diversity, physiology) and ethological (gregarism, migrations, etc.) characteristics of fish communities have a particular influence on the echoes recorded, which become characteristical and thus are able to discriminate acoustically several populations. An example of application of the method is given for Eastern Venezuela, and the use of the results for mapping, evaluation and stratification of acoustic data is discussed.

Daytime 38 kHz mean volume backscattering strength (MVBS) of small pelagic fish in the Gulf ofSt. Lawrence was recorded at 60-m intervals along a regular network of 11 parallel transects in August 1991. Profiles of temperature, salinity, density and transmittance were collected on a grid of 44 punctual stations along the transects. The MVBS was contagiously distributed and autocorrelated for al1 transects. No significant monotone relationship with any of the environmental variables was evidenced, except that the highest concentrations of fish were located over shallow bottoms. The anisotropic variogram of MVBS indicated three different scales of spatial structures: an unresolved small-scale variability, which accounted for only 7.7% of variance, and two structure components of along-transect scales of ~500 m (~40%) and 15 km (~50%). The theoretical variograms deduced for larger echo-integration units (EIUs) than the recorded 60-m unit, provided a good fit to the corresponding experimental variograms up to EIU of 1920-m, when the disturbing effect of outliers on the variograms was removed. Global estimates and maps produced with 240-m and 1920-m EIU data were almost identical, showing that no significant loss of information was caused by increasing the EIU. Asymmetric confidence intervals of the global estimates are obtained using two variograms, one controlled for the effect of the outliers, the other one not controlled. In an effort to take the relation of MVBS with the bottom depth into account, four differentkriging methods were compared: ordinary kriging, kriging with the bottom depth or its indicator-transform as external drift, and 3-D ordinary kriging with depth as the third dimension. Because of the weak relation of MVBS with the bottom depth, ordinary kriging had the best performance. Results are discussed in the context of applications of geostatistics to estimation and mapping of acoustic survey data.