BIASES ASSOCIATED WITH VISUAL CENSUSES OF REEF FISHES IN THE EASTERN TROPICAL PACIFIC: IMPLICATIONS OF USING SINGLE VS MIXED-SIZE TRANSECTS, OBSERVER EXPERIENCE, AND FISH SIZE ESTIMATION.

Abstract

Visual censuses on belt transects implemented to assess reef fish populations in the Eastern Tropical Pacific (ETP) vary greatly in the size of sampling units and census protocols. To examine the effects of such variation on estimates of species richness and density, and to help identify an appropriate (accurate, precise and efficient) protocol, we compared two belt transect methods: a single-size transect (one 30 x 2 m band) and a mixed-size transect (two bands, a 50 x 5 m band for large, mobile fishes, and a 50 x 1 m band for small or cryptic fishes). Three observers with different experience conducted 72 visual censuses on three fixed transects at El Arrecife coral formation in Malpelo Island, Colombia, following a factorial design to evaluate the effect of methods, observers and fish size estimation on the accuracy and precision of species richness and density estimates. Additionally, we examined the efficiency (cost in time in relation to the accuracy and precision) of each method. The Mixed Transect (MT) yielded more precise estimates of population and community parameters and more accurate estimates of species richness. However, it had a high implementation cost (25 min/census) and consistently underestimated total density and the density of species recorded in the 50 x 5 m transect (probably due to its greater band width and associated decrease in species detectability). The Single Transect (ST) was more cost-efficient (14 min/census) and produced more accurate estimates of total and specific density for most species (except cryptic or small ones). Both methods were biased by the estimation of fish size during the censuses and by observer inexperience, so they should be implemented under oceanographic conditions that facilitate carrying out visual censuses and by divers with prior training. Alternatively, random sampling can be performed independently of visual censuses to estimate size frequency distributions without affecting estimates of abundance and species richness. Given the generally poor visibility in the ETP and insufficient funding for research and monitoring projects, the ST protocol is more suitable for conducting surveys of multiple species. In addition to producing more accurate estimates of total and specific density for most species, the greater cost-efficiency of STs allows to increase the number of replicates and thus to improve the estimate of total species richness and the precision of all population and community estimates. However, due to the greater accuracy of the MT to assess total species richness and density of small and cryptic species, an alternative would be to use a MT of at least 30 m length, with a maximum width of 2 m for censusing large and mobile species, and of 1 m for small and cryptic species.