Martin has presented a graph based on probably
NUMBER OF OBSERVATIONS which is not a reflection of abundance. The data
presented look to be correct but the raw nature of the data make them unsuitable
for use in that manner. I point out that from the very start of reporting
on the GBS results (in the COG ABR of 1981-82 in 1983), it was realised that
observer activity is irregular through the year and between years and that this
must be included in any presentation of results. Therefore the bird data can
only accurately be represented by being shown as numbers (sum of counts for the
A statistic or tally of observations for the R statistic) both divided by
observer weeks. I find it very frustrating and
hard to comprehend why this principle appears to be abandoned now. If there
is a good reason for making this change, I welcome an explanation of this,
before presenting data.
Martin's graph is probably only a count of
observations and not of numbers of birds (as in does not include anything
about how many birds were recorded, although in this species, few
observations are of more than one bird, so there is not much difference between
the A & R statistic, as shown by that G = 1.09). More importantly
the graph is substantially influenced by number of observer weeks. So that
it is almost as much a reflection of observer activity, as it is about presence
of the birds. This is clearly shown by the obvious dip in numbers at week 52.
This is entirely due to few observers present on the last week of the year
and nothing whatever due to fewer birds present on that week. Slightly less
obvious is the another bit of wonkiness is due to the influence of Easter, being
on different weeks each year. Similar comments for other irregularities.
Thus the graph should treated as showing a large
component of the variability of observer activity as well as bird data.
The graph showing observer activity through the
year is in Figure 6 of the GBS Report and Martin's graph can best be credibly
considered through direct comparison with that graph (or the same information
for the 26 years). Doing so is a little awkward though because Martin's graph
shows the year as Jan to Dec, rather than the way the GBS year exists being July
to June. I believe that bit of intellectual effort should not be necessary.
Martin's graph should divide by the number of observer weeks, before presenting
it to the wider audience because it is potentially misleading.
The text from the GBS Report follows:
Olive-backed Oriole Oriolus sagittatus
A woodland and forest inhabitant that is a summer migrant to our area. It
passes through the suburbs on arrival and can be quite vocal at that time. It is
an accomplished vocal mimic. On most years taken individually, there is a dip in
the warmer months' abundance here, probably as the birds move away from the
suburban areas to breed. However the timing of this is inconsistent. So when all
years are combined on one graph, it does not show. It is rare here from
May to July, then birds start to arrive in August and increase to a
peak in October then decline fairly smoothly through February, March & April
when passing through the city. They are probably less conspicuous then. Although
there is some random variation, overall the long-term abundance of this species
has increased. Only three records with activities at nest, they are from mid
November to early December. Also dependent young from early November to early
February. Graphs on page: 102, Rank: 38, Breeding Rank: 42,
A = 0.05556, F = 41.80%, W = 34.0,
R = 5.119%, G = 1.09.
Philip
|