A Design for Estimating Visibility Bias in Aerial Surveys

In aerial census data, is present because the failure to observe all animals can result in severely biased population density estimates. Assuming that the aircraft can accommodate two observers situated on the same side, we present a design for estimating the magnitude of visibility bias, i.e., the probability of not observing an animal. The magnitude of visibility bias depends strongly on the propensity for animals to occur in groups and, thus, the design considers each possible group size separately. Two related methods of inference for the visibility bias parameters and the population total are described. We also discuss the application of the basic design and the associated survey methodology to an aerial survey of white-tailed deer in west-central Alberta.