Wednesday, June 5, 2013

Bird populations increase during weekends

Yes, the populations of birds increases drastically during weekends. This apparent bias in recording bird sightings during weekends in citizen science databases has been quantified for Europe and North America by Courter et al. They see a decreasing effect of the weekend effect over time for the bird species that they analyzed.

MigrantWatch is a 5 year old citizen science initiative in India that has recorded as many as 22,622 reports as of today. Although the web interface started in 2007, the dataset available for download does have 63 observations that are older, going all the way back to 1910 (Whitehead, CHT. 1911. JBNHS 21).

To get started with analyzing this dataset, it might be interesting to see how the weekend bias has changed over the years. Since, many of these reports are filed by amateur birders it should definitely show a weekend bias. In-fact Dr Jayant Wadatkar fondly remembers a weekend trip to Malkhed Reservoir when he sighted a Common Shelduck for the first time in Maharashtra. While such anecdotal information and common sense would suggest that the weekend effect is real, quantifying its magnitude might be more useful.

So lets tabulate it using the following lines:

 read.csv("migrantwatch_reports.csv",row.names=NULL)->M  
 colnames(M)<-c("Species","Location name","City","State","Reporter","Date","Sighting type","Observation frequency","Start date","On behalf of","Latitude","Longitude")  
 data.frame(Day=format(as.Date(M$Date), format="%A"),Year=format(as.Date(M$Date), format="%Y"))->WD  
 table(WD)  

Tabulation of all observations per day of week from 2007 onward

Day2007200820092010201120122013Total
Monday481862172262496622771865
Tuesday632492082282785623201908
Wednesday742281872363638072912186
Thursday642282522852707223132134
Friday492443003063066465082359
Saturday12349147761464216155874549
Sunday216848835895954247713337558
637247424762790306274913629

The weekend (Sunday more than Saturday) effect does seem rather clear just by eye. Observations for each of the weekdays ranges from 8 to 10% while Saturday has 20% and Sunday has more than 30% of the observations.

While this will obviously affect estimates of first sighting and last sighting for a season in the short term, over time this effect should not affect the larger patterns. However, one has to definitely check the effect of the day of week on all results that are obtained.

Due to the nature of the data collection effort the weekend effect is something that one has to live with.



2 comments:

MigrantWatch said...

Very interesting, Nagarjun! The number of records will certainly be influenced by when birders go birding -- not just according to day of the week, but also seasonally, and perhaps also according to particular holidays or vacation periods.

The database also contains a field about how frequently the reporter visits that particular location. When people have specified "daily" for this frequency, the location of observation is likely to be around their home or office. Perhaps if you filter for such observations, the weekend bias disappears?

Thanks for using MigrantWatch data. We encourage the wide use of the raw data, so do spread the word!

Happy birding,
Suhel (for MigrantWatch)

Author said...

Thank you for the feedback Suhel.

What you said about "daily" observations is correct and the same about holidays.

http://nagarjunv.blogspot.se/2013/06/importance-of-recording-observation.html