Addressing sampling bias in counting forest birds: a West African case study

No one bird survey technique is perfect. Either the assumptions made by a technique are difficult to meet in the field, or there are biases due to the observer, the birds themselves, the landscape, or the nature of the habitat. These include survey effort, time of day, time of year, edge effects, vegetation structure, and variation in detectability functions. Biases in methodologies are only a problem if they are systematic with respect to the variables that are being tested. If, however, biases are randomly distributed with respect to the variables under consideration, they cannot affect a positive result, although a negative result may occur because of the noise they may introduce into the analysis. Therefore, it is crucial to determine and minimise biases in any study. We use extensive line transect census data from a study of the effects of habitat fragmentation on forest birds in Nigeria to illustrate the range of possible biases and their effects, and then how these may be accounted for. Asymptote a...