OBSERVER EFFECTS AND TRAINING IN UNDERWATER VISUAL SURVEYS OF REEF FISHES

Visual survey techniques are used widely to estimate abundances of tar get organisms in terrestrial and aquatic environments. There are a num ber of methodological 'errors' in almost all applications of visual su rveys. Given the dependence of all visual survey data on the skill and technique of the observer, one potentially important source of imprec ision and/or bias is variation among and within observers. In studies involving large amounts of fieldwork over great geographic range and m any years, it is inevitable that observers will change from place to p lace and through time at any single site as they are replaced or gain experience. We present the results of 3 observer training/calibration exercises that indicate that observational studies in which multiple o bservers must be employed may be subject to considerable observer-rela ted biases and imprecision. We found that careful training and calibra tion of observers ameliorated such effects for most taxa, but non-triv ial levels of bias for some taxa and imprecision in estimates for seve ral taxa remained even after thorough training. It is essential that t he influence of observer bias and imprecision be well documented in mu lti-observer monitoring studies, so that (spurious) patterns related t o differences among observers can be distinguished from real spatial o r temporal patterns in the environment.