similar numbers in different sampling units (as in Habitat B in Figure 2.4),
then counts from different squares are likely to be similar and the estimates of
population relatively precise. If a bird has a more clumped distribution, giving
lots of variation between sampling units (as in Habitat A), then counts from
different squares are likely to be dissimilar and have lower precision. Note that
differences in the ability of the observers to make the counts can also lead to high
variance even when the birds are actually evenly distributed.
Biasoccurs when our estimates are either systematically larger or smaller than
the true value. Put another way, inaccuracy is brought about by bias, which can
arise from a poor sampling strategy (e.g. by only surveying the best areas) or an
inappropriate field method (e.g. by counting around midday when a species
is most active in the morning), or a combination of factors. A whole range of
factors could lead to bias, for example, the field method, effort and speed of
surveying, the habitat, the bird species and their density, the time of day, the
season of the year, the weather conditions, double counting, the observer’s skills,
etc. The challenge is, first, to recognize all the potential sources of bias and,
second, to standardize survey methods and improve standards where appro-
priate, to reduce bias as much as possible. That said, bias is an unpleasant and
often unavoidable fact—and surveys should always consider the likely sources of
bias and how they might influence the findings. We should never assume that our
survey is free of bias.
26 |Bird census and survey techniques
High variance
Low varianceHabitat AHabitat BFig. 2.4An illustration of uneven variance between two habitats. The filled symbols
represent birds.