Biases in estimating population size from an aerial census: a case study in the Kruger National Park, South Africa

In the Kruger National Park (KNP), South Africa, aerial census data for approximately 15 herbivore species were collected from 1981 to 1993 using a total area count, strip transect method. No estimates of bias or precision error were obtained for the census data. Visibility bias, however, has been shown to be a primary source of error in aerial census data collected using methods similar to the KNP's. Starfield argues that a pragmatic modelling approach can be used to understand the importance of uncollected data and improve data collection strategies. Following this pragmatic approach, we develop a simple, deterministic model to estimate the potential range of bias in the KNP census data. Sources of visibility bias considered in our model include undercounting detected herds and failing to detect small herds. We apply the model to data collected for impala, zebra, wildebeest and waterbuck, because these species represent a range of potential censusing challenges. The model suggests that visibility bias represents a major source of error in the KNP census data. In particular, the model indicates that visibility bias may confound comparisons among species or comparisons of a particular species' abundance among years, under different environmental conditions, or in different habitat types.