Tuesday, 14 February 2012

Noisy Miners in the COG Garden Bird Survey

As Coordinator of the COG Garden Bird Survey I was recently asked to provide information about the changes in the information about Noisy Miners collected in the Survey. 

There are a number of ways of presenting the summary results of the GBS of which the two most commonly used are:
  • A (for Abundance): the aggregate number of birds of a given species reported each active survey site-week; and
  • F (for Frequency): the proportion of sites in a given period - typically a year - in which the species was reported at least once.

The Chart below shows these two summary statistics for each of the 30 years of the GBS. (Note that each year begins on 3 July and Year 1 was 1981-82.)  As usual click on the image to get a larger version.
The smooth lines are the 4th order polynomials delivered by EXCEL, which I find a convenient way of smoothing out the 'noise' in the series.  In both cases the values of R2 suggest the smoothed lines are a good fit to the raw data.   The broad picture offered by both series is similar: a reasonably constant level of reporting until about year 19 (1999-2000) followed by a strong increase since then.

Constant Sites
An issue of concern was that this pattern was in some way influenced by changes in the composition of the Survey panel in the later years.  Since the sites are self-selecting (ie anyone who wishes to participate is welcome to do so) it might be possible that something had happened to change the nature of the set of sites.

A simple way of compensating for such effects is to use a subset of sites which have participated in the survey for a large number of years.  29 sites have been in the GBS for 15 of the 30 years: this seemed like a 'reasonable' way of ensuring that changes in the results should not be unduly affected by changes in the panel.  29 sites met the criterion of 15 or more years of contribution. typically they contributed between 25% and 30% of the sites active in a year, although this was higher (above 50%) when the overall survey participation went though a low patch in the late 1990s.

I calculated both numerators and denominators for A and F (as defined above) for the subset of sites active for 15 years or more.  The next two charts compare the values of these two summary statistics for all sites (as shown above) with the equivalent statistics for the >15 year sites.

Not surprisingly the graphs for the >15 year series are a little less stable than for the wider sample.  However allowing for the possible differences between the two series the degree of agreement is very good.  I therefore conclude that the pattern shown in the first Chart is not an effect of sample composition, but due to some other effect.  I will leave it to others to suggest causes for that.

Changes in Flock Size
Reviewing the first chart in this post suggest that the value of A is rising faster than the value of F.  This implies that flock size s increasing as well as the frequency of observing the species.  Within the GBS Summary statistics, G (Group size) shows the average flock size by year.
This chart suggests that group size has been increasing gradually over the last few years.  Knowing that the species is usually encountered in modest sized flocks, but in Winter the groups can aggregate to quite large flocks I charted the proportion of total flocks with 10+ birds.  While the result is rather unstable in recent years the smoothing polynomial shows a reasonable increase in recent years (particularly during the drought years).
A further chart was compiled showing the proportion of flocks according to flock size during the the most recent  7 years and for earlier years.

In broad term this shows a higher proportion of very small groups (1,2 or 3 birds) in the early period, approximately similar distributions for the mid-size flocks and a lower proportion for the larger flocks (8+ birds).

Seasonality
I have also examined the seasonality of reports over these two periods. 
The most consistent period of higher reports for the more recent period appears to be in Autumn - weeks 15 (beginning April 9) to 21 (beginning 21 May).  This would seem to coincide with a migration out of the high country following the onset of colder weather.  However HANZAB rates the species as sedentary, with most suggestions of seasonal movement dating from before 1950.  Possibly the GBS data reflects the formation of larger, more raucous and thus more readily observed groups at this time?

No comments: