Jump to content
Ornithology Exchange (brought to you by the Ornithological Council)

Problems with Return Track

Recommended Posts

Hi, all

I'm using FLightR 0.4.9 to track migration of shorebirds. Tags were deployed in Alaska and recovered at the same site a year later. FLightR gives me nice-looking tracks for the breeding season in Alaska, fall migration to Chile, and the non-breeding season in Chile. But the track for spring migration is consistently problematic  -- instead of returning north to Alaska, it heads south toward Antarctica.

Has anyone experienced something similar? Any ideas for what I can do?  

Calibration periods are from the study site (lon = -150.725, lat = 61.272): one from the beginning of the study (shortly after tag deployment) and another from the end (shortly before tag recovery). Both verified by plot_slopes_by_location.

> Calibration.periods <- data.frame(calibration.start=as.POSIXct(c("2009-05-17", "2010-05-10")), 
                         calibration.stop=as.POSIXct(c("2009-06-15", NA)),lon=-150.725, lat=61.272)
> print(Calibration.periods)
  calibration.start calibration.stop      lon    lat
1        2009-05-17       2009-06-15 -150.725 61.272
2        2010-05-10             <NA> -150.725 61.272

Calibrating from pre-deployment data instead yields similar results. For the particle filter run I've tried known.last=TRUE and known.last=FALSE. Changing behavioral masks, outlier check, nParticles, etc. also do not seem to help. 

> all.in <- make.prerun.object(Proc.data, Grid, start=c(-150.725, 61.272), Calibration=Calibration)
> nParticles=1e6
> Result <- run.particle.filter(all.in, threads=-1, nParticles=nParticles, known.last=TRUE, check.outliers=TRUE)

Since I'm using the correct lon/lat at the end of the track to calibrate, why might the output show my end coordinates so far away from there?


Link to comment
Share on other sites

  Hi Jenny,

 1. You could try increasing the distance they are allowed to fly. The maximum distance can be fixed in particle.filter and the mean and sd in make.prerun.object.

2. Try running first without any spatial constrains. It sometimes happen that particle filter just cannot find a proper solution if bir migrated to far every day for a long period.

  Hope this helps,


Link to comment
Share on other sites

Thanks for your help, Eldar! Distance does seem to be key. Increasing max distance in particle.filter has improved the return track.

Now working on adjusting mean and sd. Forgive a beginner question, but is there an easy way to estimate appropriate values? I have the result of a particle.filter run that produced an OK-looking track, but not quite sure what I should use. Should I include all movements in my estimates? These are birds that remain in a relatively small area for most of the year, but travel long distances  quickly during migration... 

Link to comment
Share on other sites


This topic is now archived and is closed to further replies.

  • Create New...