marginal to the regular avifauna. The same thing happens in a smaller study area
where there is a risk that later additions to a list might be irregular transients.
To demonstrate that a list is virtually complete, it is necessary to show that the
rate of accumulation of new species with further effort has reduced to an accept-
ably small number. By this point, the number of species recorded only once will
have fallen to a very low level. In the field, it is helpful to have a list of the likely
or possible species so as to focus attention on checking for the missing ones that
might occur. Do you know these species well enough to have a good chance of
picking them up? Have you looked in the best habitats at the right time of day
and season to find them if they were there?
Ideally, species number would be compared across plots that have either
received equal effort or sufficient to obtain a near full list for each. It is possible
to make a retrospective estimate of the number of species for a fixed effort that
is less than the total put in by resampling the data. In this way, all plots can be
compared at a standardized effort level of that which received the least but this is
rather wasteful of data. Alternatively, it is possible to estimate the number of
missing species from the total numbers observed and the numbers found only
once or twice (see Colwell and Coddington 1994 and Boulinier et al. 1998).
There are programs at http://www.mbr-pwrc.usgs.gov/software/comdyn.html
A much used study design looks for the effects of fragmentation and habitat
modification by comparing species richness across a range of sites. Species lists
are accumulated, usually by timed visits to study plots, which might vary in size
by several orders of magnitude from a few hectares up to tens or hundreds of
square kilometers. It might take only a few hours to estimate the species richness
in a small plot but considerably longer to complete the list in a large plot. It is
tempting to under sample the smaller plots when the species list is slow to grow
and put more effort into the larger ones where new finds continue to look likely.
A sound quantification requires the demonstration that the species lists for all
plots are comparably complete. Ideally, each would have received several visits
beyond the point where the list ceases to grow. This can be done with a stopping
rule such as stop when the number of species seen only once is less than or equal
to the number recorded twice.
1.8 Conclusion
The development of bird survey and census methods really got going about 40 years
ago. There was a major conference on the subject 20 years ago (Ralph and Scott
1981). Reading the proceedings, one could be forgiven for concluding that
no method produces consistent and reliable results. The list of potential biases
Conclusion| 11