Friday, 30 June 2017

On Owls and Owling

Some years ago Birds Australia published research that appeared to show that Boobooks were threatened in Australia.  Knowing, from McComas Taylor's work on the Atlas of ACT birds that it is crucial to conduct surveys at night to count this species I enquired if the drop between the 1st and 2nd Atlasses could be due to there being less Owling (nocturnal activity by birders) rather than a decrease in owls.  To my total expectation I never got an answer to my question.

There is now a fair body of information around from eBird that is publicly available so I thought I would have a look at this to see what it could say about the less uncommon owls in the Canberra area.  It turned out to be a very interesting exercise in many ways.  There are quite a few graphs in what follows, but hopefully the text will be enough to keep the interest going for the cartophobic.

The three owls I am examining are Southern Boobook (hereafter Boobook), Powerful Owl (Powl) and Eastern Barn Owl (Barn Owl).  I will look at each of these in three ways in what follows and conclude with some commentary about eBird and the entertainments available from downloading it!

Which Month are they seen in?

It is thought that many of the records relate to hearing birds calling to attract mates so it could be expected that the number of reports will vary by month.  That is shown in the following chart: note that to enable easy comparison of the shape of the lines I have multiplied the number of Barn Owl records by 3.
According to HANZAB V4 Powl breeds in Winter,  Boobook in September - November and Barn Owl throughout the year (influenced by mouse plagues).  In view of those comments it is a little surprising that all three series show peaks in April-May and September - October.  My immediate thought is that those are the best months for birding- like Baby Bear's porridge, not too hot and not too cold!)  I shall return to this later.

What o'clock are they observed?

This next line of enquiry looks at the time of day when the birds are reported.
The pattern for Boobook and Barn Owl is very similar (with peaks in the evening and early-mid morning) while that for Powl is rather different with a peak late morning to afternoon.  They all agree that there are few records in the middle of the night, suggesting that birders, if they aren't Owl specialists, still need their beauty sleep.  This issue will be dealt with by using reporting rates in the third section of this report.

With regard to the Powerful Owl situation I suggest that this is because the birds are sufficiently rare to be of great interest to birders and when one is found roosting lotsa people will turn up to tick it at a time suitable to them.  By way of examples:
  • Of the 137 eBird records for Powl in the ACT 114 (83%) relate to a bird that roosted near the Turner Bowling Club for ~18 months.
  • The most frequently reported site for the species is "Mt Coot-tha Reserve--JC Slaughter Falls" from which a number of reports have been made by a friend from Brisbane.  He has advised that a pair nest here and  "Most of our eBird lists will be to specifically try and locate them."
I also looked at the combination of time of year and hour of the day for Boobooks.  The chart was very messy if plotted for months so I combined months into the 4 traditional seasons.
Winter obviously has a much lower number of records than the other three seasons.  The pattern for those other seasons is quite similar with a tendency for the peak calling times to be later in the morning and earlier in the afternoon according to day length.  (In other words the birds seem to be active around sunrise and sunset; this is consistent with comments in HANZAB.)

Reporting rates

I have referred in a couple of places above that the patterns of reports seem to be influenced more by the activities of the birders than the birds.  So I decided to create a measure that showed the proportion of records that included Boobooks.  On consulting eBird Central the only way of getting a count of overall numbers of checklists was to download records for all species and then calculate the number of checklists.  

The next three paragraphs - in blue - can be skipped by those not trying to calculate reporting rates for themselves

The core problem with this is the size of files that results.  I knew that an "all species - all years" file for Australia would be huge so thought I would work with a file of this nature for NSW.  This was about 80Mb which expanded to nearly 1Gb when unzipped.  None of my programs seemed keen on processing this (ACCESS had the best effort but kept refusing to handle the crucial date field)! 

So I tried a file for all NSW species restricting it to 2015 and 2016.  This was still 50Mb when zipped but I was eventually able to load it to EXCEL from where I deleted many of the variables not of interest to me in working out how many checklists were lodged each month and each hour.  There are about 770,000 records in this file! 

My next step was to extract Month from the Date variable and Hour from the Starting Time variable and use those as denominators in calculating reporting rates. There was also considerable entertainment in stripping checklist number out of a field  GLOBAL UNIQUE IDENTIFIER of the form: URN:CornellLabOfOrnithology:EBIRD:OBS220444969 where I think the checklist number is the element in bold red.  However it was done, giving me 31472 checklists (thus an average of 24.4 species per checklist) to work with.

OK so now I can calculate reporting rates as the percentage of NSW checklists that included Boobook and classify these for hours and months.  Here is the chart for hours.
As expected it shows that Boobook records become very important before 6am and after 9pm. 

Let us now have a look at reporting rates for NSW Boobooks by Months.
Interestingly this chart still shows a pronounced low in Winter.  The (relatively) few birders who go out in Winter seem to have more trouble than usual in locating Boobooks!  This possibly reflects what appears to be - from my reading of the species account in HANZAB - a low rate of vocalisation in Winter, building up to a noisier lifestyle prior to egg laying.


An interesting exercise in I think demonstrating the importance of nocturnal records in assessing the status of Boobooks.  There is also a suggestion that careful thought needs to be given to the impact of 'heard' records, and seasonality, in coming to such conclusions.  

For Powerful Owl any analysis of citizen science records needs to take particular note of the likelihood than a large increase in records reflects one bird recorded many times.  (In that regard I'd mention that despite the best efforts of eBird moderators to use standard names for sites at least 5 location names refer to the area around the Bowling Club and thus the same bird.)

The reporting rate analysis for Boobook was interesting to work on but took close to a day of very frustrating effort.  Obviously this would be less of a problem for a Serious User, with industrial computing power, decent internet access, and someone else paying for the downloads.   I'm not sure what the answer is but suggest that in view of the importance of the number of checklists in such exercises it would be useful for eBird to develop a standard "product" (pass the marketing jargon barf-box please) with checklist counts to sub-national level x month (while hour was needed in this exercise, as a generalisation I'd rate it as nice to know rather than essential).  

No comments: