2.2 Sampling strategies
We saw in the previous section that, if we are to obtain an unbiased measure of bird
abundance (e.g. an estimate of absolute or relative population size), we will often
need to count birds in a number of sampling units that are representative of the area
within the survey boundaries. This raises two important questions; how many
sampling units should we visit to count birds? And, crucially, which ones?
2.2.1 How many sampling units?
As we have seen, the larger the sample size (number of areas and hence birds
counted) the more precise our estimate. Sample size will therefore depend largely
on the reliability we want to place in our estimate. If we want a very precise
estimate, we need to have a larger sample of sites than if we just want a good
approximation. Statistical methods, requiring the collection of some pilot data, are
available for calculating sample sizes necessary to achieve predetermined levels of
precision (Snedecor and Cochran 1980). In the real world, however, our sample
sizes are generally influenced by financial and human resources, and, as these are
generally low, we will rarely be at risk of having sample sizes that are much higher
then we actually need. Instead, we need to ask ourselves whether our sample size
will be sufficient to meet the objectives that we set ourselves at the outset.
2.2.2 Which sampling units to count?
Next, we need to determine which sampling units, out of all those available,
should be visited. In other words, what is our sampling strategy? This is probably
the most critical decision in a sample survey, as failure to use an appropriate
sampling strategy could invalidate the results. Only when we are certain that
our sampling strategy is appropriate should we start to think about how we will
actually count the birds when we get into the field.
There is a tendency for fieldworkers to visit areas they expect to be good for their
target species or for their particular study. Free choiceof this kind can lead to a bias
toward higher quality sites, or particular types of site. Remember that our sample
must be representative of the whole area of interest if we are to extrapolate the
results to areas that are not visited. So how can we select our sample without fall-
ing into this trap? The most frequently used methods, and the best, are random
samplingandregular sampling. A definition of truly random sampling is that each
sampling unit has an exactlyequal chance of being selected. Contrast this to free
choice, where better areas are far more likely to be selected than less good areas.
Sampling randomly is not as straightforward as it might seem. One might
think that closing ones eyes and sticking a pin in a map would be random, but it
Sampling strategies| 27