Travelling records in eBird

I was recently in a discussion with some folk from 2 Birdlife Australia (BLA) groups about the use of citizen science data to monitor the recovery of birds from the Black Summer fires.  I suggested that as well as the data contained in Birdata (which appeared to be biased towards the coastal areas) some thought could be given to supplementing this with data from eBird.  One participant - with a scientific thrust to his interests - has subsequently commented about having some doubts about the value of the data from the eBird 'Travelling' protocol.

To date I hadn't really focused on such an analysis,  Before the cries of "Shime, shime" become too loud I point out that my interest has been more to compile monthly checklists of species seen in a fairly large area.  Thus as long as it appears that (for example) the records were all within the Cape Howe Wilderness, which is completely within my study area, the fact that some of the lists were ~80 km doesn't fuss me.  (As an aside, given that some parts of this area would have been burnt and others weren't, the records in this checklist are not useful for a fire-impact study.)

 

Thus I thought I would have a look at the eBird data for East Gippsland Shire to assess how the data could be filtered to be useful.  By way of context up to 31/12/2019, eBird holds 222,739 species records for the shire. contained in 15,958 checklists.  A number of the checklists have been shared between members of groups which introduces some amount of duplication which might be important for some studies: using the group identifies code to eliminate this gives 13,199 groups.

In what follows I have mainly tallied species records, which is somewhat imprecise due to the duplication from sharing, but I think the overall picture is satisfactory for the purpose.  There are a number of protocols available within eBird (quite a few of which are not relevant to Australia.  This first chart shows the number of species records held against each protocol.

The red bars indicate protocols which IMHO can be directly mapped to a BLA protocol.  It is possible that the records from within a list compiled while stationary might include birds seen but outside a notional 2 Ha site (which could be [eg] a circle radius of ~80m, or a rectangle 200m x 100m) but I don't believe that to be a huge problem.  Note that if a route totals <10 km it must by definition fit within a 5 km radius.  Indeed, many of the routes with a length >10 km could fit within such a radius but I am being conservative.

In total I estimate there are at least 136,332 species records which meet those protocols.

Obviously, the larger the area the greater the diversity included within it.  Ebird advise that a new list should be started where habitat changes, but that is frankly going to be obeyed in the breach.  A citizen scientist (aka "birder") walking along a 4km track that includes 2 km of heath; 1.5 km of woodland and 0.5km of beach will not provide three lists whether using eBird or Birdata.  This second chart compares three lengths of walk in terms of the proportion of lists split to classes of number of species.  (An example may help: 68% of trips <0.5 km have 10 or less species: the equivalent % are 30% for 0.5 -5 km and 13% for 5-10 km.)

EBird recommendations are also to submit a new list if the distance travelled is greater than 8 km.  That is followed in most cases but there are 15,339 species records (7% of total records) on sheets covering >10 km.  I believe that to be simply part of the nature of birding - people want a simple life and don't want to stop and spend 5 minutes setting up a new site.   As noted earlier for my purposes that is quite OK but for analysts attempting a more structured approach they may wish to filter out these records.  

For curiosity I looked closely at the 14 checklists with a length of 50km or more.  10 of the lists reflected group outings so there are 8 'outings' covered by these lists.  

• 4 lists are by 2 groups of overseas birders (presumably more intent on building their world list than contributing to detailed science - been there, done that);
• 3 lists are a group of three birders (all reputable names) exploring the Cape Howe Wilderness (see preamble)
• 3 more lists are the first ever pelagic into the area off Gabo Island;
• the final 4 lists are simply people having a pleasant drive with birds!

It is difficult to see how these records could be incorporated in an analysis based on the standard BLA protocols.  However if the analysis was looking at "species of interest" it may well be of interest to include records of those species even if the area of occurrence is more generalised than desired..

Continuing the theme of analysis for "species of interest"  there would seem to be some merit in also including records for those species compiled using the incidentals protocol (and/or historical records).  By way of example records under the incidental protocol include 3 records each of Regent Honeyeater (5 records in total for East Gippsland) and Australasian Bittern (no other records for East Gippsland).