different habitats for a radio-tagged or otherwise individually identifiable bird,
which has been followed and its location recorded at intervals.
Deciding which and how much of each habitat is available is an important
part of any analysis. For example, foraging habitats that are a long way from the
nearest roosting or nesting sites may not really be available, or at least not as avail-
able as those near those habitats, and this can lead to bias if the habitat also varies
with distance. A solution to this problem is to look at selection within areas that
are carefully defined so that they can be regarded as equally available (e.g. within
a certain distance). Alternatively the distance to the nest site or roost can be con-
sidered in the analysis. This approach allows different kinds of habitat preference
to be examined separately.
A simple way to analyze this type of data is to calculate the density of records
(numbers of birds or bird locations) in each habitat. If the records can be regarded
as statistically independent from one another, a chi-squared test can be used to
compare the observed distribution of records across habitats with that expected
if record density was the same in all habitats. However, records are often not
mutually independent and assessing preference becomes more complicated if
there are several study areas, several survey dates or if a radio-tracking study
covers several individually identifiable birds. Pooling bird count and habitat area
data from several study areas or time periods can give misleading results if the
relative areas of the different habitats vary amongst study areas or over time. It is
also the case that the values of often-used measures of preference or selection,
such as the forage ratio, Ivlev’s index and Jacob’s index, change if the relative areas
of habitats available differ, even if the ratios of the densities of bird records in the
habitats remains the same. This is clearly unsatisfactory and these indices are not
recommended when it is envisaged that results for several areas or time periods
with varying areas of habitats will be analyzed.
A measure of preference that does not suffer from this defect is the Bi1index of
Manly et al. (1972). The calculation of this index involves first dividing the pro-
portion of bird records in a habitat by the proportion of the available area in the
habitat. This is done separately for each study area or time period. This is the
forage ratio. The forage ratio for any particular habitat is then divided by the sum
of the forage ratios for all habitats to give an index that will have the same value
in different areas or surveys provided that the birds maintain the ratios of their
densities across habitats at the same values.
Statistical testing could be done on the values of Manly’s index from independent
areas or surveys to examine the degree to which habitat preference was consistent
across areas, or survey times. However, a more satisfactory approach to the analysis
of data from many individual birds, study areas, or times is to use compositional
Quantifying habitat selection| 261