Simulation d'un plan d'expérience en océanographie: Effets de la dispersion des organismes sur l'erreur d'échantillonnage

Abstract

Seven spatial dispersions of organisms, reflecting some constraints of various physical and biological factors of the pelagic environment were simulated on computer. The relationships between the sampling methods (tow length, net size), the pelagic population dispersion patterns, and the sampling error were investigated by computer simulation. Two categories of dispersions were identified: 1) random or relatively uniform dispersions of organisms (including Poisson, gradient, and negative contagous dispersions), characterized by an index of dispersion s 2 m < 1.5 ; and 2) overdispersed distributions of swarming or clustering organisms (including positive contagion and centric aggregation dispersions), characterized by a value of s 2 m > 1.5 . The origin and size of the sampling errors as deduced from the analysis of variance change, considerably from one group of dispersions to the other. In the first case, the net size was dominant, whereas in the second case the tow length and patch size effects are more critical. In general, the precision of estimations decreases as the over-dispersion of organisms increases. The confidence limits, for a simple estimation of abundance, vary considerably with the spatial dispersion, ranging from narrow values for the relatively uniform dispersions to very large values for the overdispersed distributions. There is an increasing imprecision in the estimate of population abundance as the dispersion coefficient s 2 m increases. Analysis of coefficients of variation indicates that more precise estimates of abundance may be obtained if a practical sampling procedure is applied to a group of specific organisms. Furthermore, it is suggested that a simulated approach, performed prior to ocean sampling, permits the establishment of the optimal sampling procedure for the species under study.