We present an application of a statistical method based on an original homogeneity test, derived by non-parametric maximum entropy techniques, that analyses ecosystems starting from their species composition. The method is applied to zooplankton data collected in different seasons from three adjacent coastal regions in the Tyrrhenian Sea (Mediterranean), with the aim of quantifying statistically significant differences in species composition which could reflect different environmental features. The analysis revealed significant differences between the study areas. The Gulf of Salerno, a wide mouth embayment open to the flushing of the oligotrophic Tyrrhenian waters, was characterized as a homogeneous area with uniform spatial patterns in copepod assemblages through the seasons. In contrast, the Gulf of Naples appeared to be fragmented into different homogeneous subsystems with boundaries shifting according to the season. The heterogeneous structure of the pelagic system in this environment is probably related to the variable local hydrography and to nutrient enrichment due to land runoff. Small, but statistically significant, differences in species composition characterizing each homogeneous subsystem were not directly related to precise ecological factors but indicated differences that did not depend on intrinsic fluctuations. Such differences in species composition could indicate stress before modifications in environmental parameters are detectable. The present method has been compared with other multivariate techniques, including cluster analysis and principal component analysis.