or every stratum, contains a fixed number of sampling units (Figure 2.8(b)). An
alternative to random sampling that gets around this problem is regularor
systematic sampling. This involves selecting the sampling units by choosing them
in a regular pattern (Figures 2.8(c) and 2.9(a)). We can use random numbers to
help us do this. If we want a 10% sample from 100 squares, we can select a ran-
dom number, say 7, then take every 10th square from a list in standard order;
7, 17, 27, 37,... , 97. Alternatively, we could simply decide to sample every
1-km square in the north-east corner of every 10-km square and so forth to
achieve a predetermined sample size. There are advantages to regular sampling
compared to a random design:
- Regular samples are easier to select—a single random number is all that is
required. - It samples evenly over the area of interest; there is ‘built-in’ stratification
that ensures that samples are taken from across the whole area of interest. - In consequence, it is often more accurate.
- It can be used to create maps and atlases.
- It is easy to understand and explain to others.
34 |Bird census and survey techniques
Fig. 2.8There are certain situations, in which a pure random sample can, by chance,
miss an important part of the study area, which could lead to serious under- or over-
estimation of a population depending on its distribution. In this example, a random
sample (a) under-samples the southeast corner of the study area. A stratified random
approach (b) could alleviate this problem by requiring a survey point in every grid
square in the study area. Similarly, a regular sample (c) overcomes this problem
because survey points are located in the center of every grid square. Here the filled
circles represent sampling units (n20) within a study area defined by the bold border.
(a) (b)???(c)