Wednesday, 2 September 2009

Getting more Superb (Fairy-wrens)

Summary
To save those of you with a busy life from reading three pages of stuff, here is the short version.

I started trying to explain why the value of A for Superb Fairy-wrens (Malurus cyaneus) drops dramatically between years 17 and 18 of the Garden Bird Survey (GBS). I didn’t explain this.

I then moved to trying to explain why the value of A for this species shows a fairly steady increase from year 18 onwards. This was more successful. It appears to be due to a number of GBS sites, active for a relatively short period with several of them joining the Survey since year 18, showing a high recording rate of a fairly large number of birds. The sites tend to be on the edge of – or outside – the urban area of Canberra-Queanbeyan.

The Detail
While doing data entry for the GBS I was surprised at the small numbers of Superb Fairy-wrens reported in some sites where I would have expected them. This led me to look at what is in the database for this species.

The key summary statistic derived from the database is denoted ‘A’, for Abundance. It is calculated as the average number of birds of a species reported per active observer week within a year. For example in the first year of the GBS observers reported 768 Superb Fairy wrens (SFW) and indicated that they were at their sites for a total of 2026 weeks. Thus A = 768/2026 = 0.3790.

Over the 27 years of the GBS for which final data has been calculated the value of A rises fairly steadily from Year 1 to Year 17 and then plummets to year 18 after which it rises again, at a slightly steeper rate than before. This is illustrated in the following graph, also showing linear trend lines for the two ‘sections of interest’, their equations and values of R2.I began to look more closely at the large drop off between years 17 (ending on 2 July 1998) and 18(ending on 2 July 2009) which is hereafter referred to as the Great Plummet. This seems to be a real change in the population, rather than an artifact of the GBS,since of the net (-)1208 difference in the number of reported birds most ((-)856 – 70%) is attributable to changes in sites reporting the species in both years. New sites added 35 birds while abandoned sites reduced the count by 387. The following table summarises these changes.
Long term research into Superb Fairy-wrens in the Australian National Botanic Gardens (ANBG) shows that numbers correlate with spring (ie August to November) rainfall. The value of A was assessed against these data from the Bureau of Meteorology for Canberra Airport. No correlation was identified: the values of R for the 27 year series was close to zero. Focusing on the period from year 18 onwards, correlating rainfall for the period August-November of year ‘n’ with the value of A for year July n to June n+1 gave a strong negative correlation (R=0.66)! Applying a lag to the values of A (using the values of rainfall in Spring for year n-1) showed a correlation coefficient of -0.04 ie no significant relationship. As well as the lack of long-term correlation, the Canberra Airport rainfall did not appear to explain the Great Plummet.

This difference in outcome is almost certainly an effect of the difference between the very systematic research in the tight area of ANBG and the far more ‘noisy’ dataset available from the GBS.

An attempt was made to examine GBS breeding data for Superb Fairy-wrens, but as is the case for many species there were relatively few (107) breeding records and 75% of these were either F (+ Fledgling, early years) or DY (Depenedent Young, later years). Possibly this paucity of data reflects SFW nests being “well concealed, close to ground and usually in live plants” (HANZAB v5 p 284) and thus hard for GBS observers to locate.

A further avenue of enquiry was whether the Superb Fairy-wrens were increasing in GBS sites as the vegetation in the gardens matured, thus providing more of the appropriate breeding sites. There is only limited habitat information provided in the GBS and certainly not enough to analyse such an enquiry on a site by site basis. However it was thought that a variable showing the time since first development of a suburb might give a proxy. Unfortunately I have not been able to find a way of analyzing the data to demonstrate any significant relationship.

A further thought was that the series may be affected by sites only in the GBS for a few years (in many cases a single year) or that less experienced birders have difficulty spotting these birds. To overcome these issues A was recalculated, restricting both the number of birds and the observer weeks to sites which have reported for a minimum of 15 years. It was then trivial to do the same for the sites in the GBS for shorter periods of time. The result is shown in the following graph. While the two series show some consistency in earlier years there appears to be a clearly higher rate of increase for the “Other” series in the years since the Great Plummet so these components of the series were examined in more detail.Looking at the linear regression equations the rate of increase is higher (shown by the ‘x’ coefficient) and more stable growth pattern (shown by the value of R2) for the ‘other sites’ series.

The next step was to examine the characteristics of ‘other’ sites contributing most to the values of A over the last 10 years of the GBS. Without being able to offer definitive proof, due to lack of GIS software (and GIS skill) it appeared that most of these sites showing large aggregate numbers of Superb Fairy-wrens were either:
* closer to the edge of the built-up area (eg Crace, Jerrabombera and Isaacs) or
* rural sites (eg Hoskinstown, Carwoola, Gundaroo).
It also appeared – on the basis of eyeballing the database records - that one of the factors was an
increased recording rate for the study species in these “high scoring sites” rather than particularly high numbers at any point in time. This tends to be supported by the review of the R-statistic below.

Of course the standard summary statistics available from the GBS are not restricted to A. The following material summarises my review of some “all sites” graphs for the variables cited. The term ‘early years’ refers to years prior to Year 18.
F: Early years show pronounced growth from approximately 50% of sites reporting the species to approximately 80% of sites. Relatively little drop between years 17 and 18. Later years show a steady rate of close to 80%.
R: Early years show pronounced growth from approximately 15% of observer weeks reporting the species to approximately 35% of observer weeks. A very pronounced drop between years 17 and 18, followed by another period of growth from 20% to low 30%s.
G: Early years show a variable Group size ranging from 2.4 to 3.6 birds, with no clear trend. After a drop between 17 and 18 later years show a somewhat fluctuating, but significant, upwards trend from 2.8 to 3.7 birds per group.

No comments: