Error analysis in the estimation of coral reef fish numbers from visual surveys

抄録

The study quantifies different sources of error in the visual censuses of 12 fish species on a fringing coral reef. Each species was censused 24 times within a period of four days. We assumed that in this narrow time window, true abundances of species were constant, and that the variation observed in the estimates was due to short-term environmental oscillations, inter-observer differences, and chance events. ANOVA revealed that the environmental variance in four species, inter-observer variance in three species, and chance variance in one species contributed most to the censusing results. In the remaining four species, the variance components were similar in magnitude. We conclude that identifying important sources of error variance, which allows for the informed correction of censusing methods, could increase the precision and comparability of density estimates.