It is possible to mix sample and census approaches within the same survey.
Thus, in some areas or habitats a census of all birds is used, for example, where
densities are high in limited geographical areas, yet in others only a sample of
areas or habitats is counted, for example, where densities are low over wide areas.
2.1.8 Sampling strategy
If we decide to undertake a sample survey, we need to be very clear about the
sampling strategy. We need to ensure that the areas in which we count are truly
representative of the area within the survey boundaries. If they are not, our final
estimate or index may be biased in an unknown manner. Strategies based on
random, random stratifiedorregular sampling(also known as systematic sampling)
are likely to be most robust. As this is such an important topic, it is outlined later.
2.1.9 Sampling unit
In tandem with our sampling strategy, we need to decide upon our sampling unit,
the bits of the whole survey area we actually count birds in. This might be a
grid square, the precise location and boundaries of which are available from maps.
The area encompassed within the survey boundary can be subdivided into a large
number of grid squares on a map, and a sample of these squares chosen at
random for survey. While this approach is simple and statistically sound, it may not
always be practical. It might be difficult to use, for example, when surveying birds
living in fragmented forest plots of variable size surrounded by farmed land.
In such circumstances, individual plots can become the sampling unit. In this case,
unlike the grid squares, the individual sampling units are likely to vary in size.
2.1.10Field methods
We now need to consider what field methodwe will use to count the birds. There
are a variety of options and the one we choose will depend upon the species or
group of species being counted, the habitats involved and the level of detail
required. For some species, it is necessary to develop specially tailored methods
(see Gilbert et al. 1998; Steinkamp et al. 2003). If we are trying to survey a num-
ber of species together, however, then we need a generic method that will encom-
pass most species well. There are two principal methods for generic or single
species surveys; mappingandtransects. These methods, plus others with specific
uses, are outlined below.
2.1.11 Accuracy, precision, and bias
The terms accuracy, precision, and biashave specific meanings when applied to
scientific data, such as bird surveys, though accuracyandprecisionare generally
Introduction| 23