Jump to content


Photo

FlightR: identifying staging areas and output map

FlightR staging areas map

  • Please log in to reply
2 replies to this topic

#1 Louise de Raad

Louise de Raad
  • General Members
  • 13 posts
  • Inverness,
  • United Kingdom

Posted 27 July 2017 - 04:48 AM

Hi, I am using FlightR on geolocator data from common sandpiper. I have pre-processed my .lux files using TwGeos package and manage to subsequently run the FlightR script without problems. I do have some questions about the results it produces and some bits of the script:

 

Do I understand correctly that a location is calculated by the slope of a twilight and that thus, in contrast to geolight, you obtain a location per sunrise and another location per sunset (rather than getting one location for each sunrise and sunset 'pair')? And that as such, excluding / deleting twilights does not mean the twilights (locations) before and after the deleted twilight are also lost?

 

In Eldar's (2015) paper (page 6 of 15) it states that for all their models, they assumed that the truncation points were a = 45 and b = 1000km. "That is, when birds initiated a movement, they could not fly less than 45km or more than 1500km in a single between-twilight interval."

  • This seems to me that 'b' should have thus been 1500. What is actually being used in the script 1000 or 1500km?
  • I take it that this means that a bird is classified as being in a movement state if it moves >= 45km from one estimated location to the next, but are geolocator data derived locations not more imprecise than that (so that the bird could be classified as in a movement state whereas the "movement" is actually due to inaccurately calculated positions)? Should this threshold therefore not be higher, e.g. >= 300km (see e.g. paper Fudickar et al 2011, but even Eldar's 2016 paper showed a 43.3 +/- 51.5km deviation from the true position)

A question about identifying staging areas using this bit of script

stationary.migration.summary(Result, prob.cutoff = 0.1, min.stay = 3)

min.stay is presumably the number of twilights, but what happens if twilights have been excluded from what could potentially be a staging area? So for example you say min.stay= 3 (which normally could mean a sunrise, sunset and another sunrise (e.g. approximately 24 hours minimum stay) but the sunset was excluded - does it then become sunrise, sunrise, sunset (i.e. approximately 36 hours minimum stay)?

 

Then a couple of questions about the outputs of the script:

 

  • In the plot of the locations (map.FLightR.ggmap(Result)), it doesn't look like all locations are plotted (I have 521 coordinates in the coords.csv file, but the map seems to only show around 70 points). What locations are plotted?
  • Which coordinates from the coords.csv file are plotted (Medianlat/Medianlons or Meanlat/Meanlons)?
  • In the coords.csv file what is the difference between Medianlat and MedianlatJ (and the same for long)?
  • In the plots of lat and long, what do the dark grey and pale grey limits show (are these first and third percentiles respectively?)

 

Sorry, a lot of questions (I saved them all up!!), I hope you can help me with.

Thanks

Louise

 

 

 



#2 Louise de Raad

Louise de Raad
  • General Members
  • 13 posts
  • Inverness,
  • United Kingdom

Posted 16 August 2017 - 04:25 AM

Does anybody have any ideas, as I'm still stuck!



#3 Eldar

Eldar
  • General Members
  • 47 posts
  • Netherlands,
  • Netherlands

Posted 18 August 2017 - 01:50 AM

  Hi Louise,

 Sorry, super busy times...

 

 

 

 

Do I understand correctly that a location is calculated by the slope of a twilight and that thus, in contrast to geolight, you obtain a location per sunrise and another location per sunset (rather than getting one location for each sunrise and sunset 'pair')? And that as such, excluding / deleting twilights does not mean the twilights (locations) before and after the deleted twilight are also lost?

 

  Correct.

 

 

 

In Eldar's (2015) paper (page 6 of 15) it states that for all their models, they assumed that the truncation points were a = 45 and b = 1000km. "That is, when birds initiated a movement, they could not fly less than 45km or more than 1500km in a single between-twilight interval."

  • This seems to me that 'b' should have thus been 1500. What is actually being used in the script 1000 or 1500km?

You can always check defaults by looking at the help file for function:

?run.particle.filter

It is currently 1500 km. You can also change it to any value you like.

 

 

  • I take it that this means that a bird is classified as being in a movement state if it moves >= 45km from one estimated location to the next, but are geolocator data derived locations not more imprecise than that (so that the bird could be classified as in a movement state whereas the "movement" is actually due to inaccurately calculated positions)? Should this threshold therefore not be higher, e.g. >= 300km (see e.g. paper Fudickar et al 2011, but even Eldar's 2016 paper showed a 43.3 +/- 51.5km deviation from the true position)

 

'a' is 45 as FLightR estimates positions on a grid and minimum distance between grid cells is about 50 km. I would not pay too mch attention to 'a' as it is only a way to estimate the model. Have never done experiments with making a higher let me know what results are if you go for it.

 

 

A question about identifying staging areas using this bit of script

stationary.migration.summary(Result, prob.cutoff = 0.1, min.stay = 3)

min.stay is presumably the number of twilights, but what happens if twilights have been excluded from what could potentially be a staging area? So for example you say min.stay= 3 (which normally could mean a sunrise, sunset and another sunrise (e.g. approximately 24 hours minimum stay) but the sunset was excluded - does it then become sunrise, sunrise, sunset (i.e. approximately 36 hours minimum stay)?

 

 

You are right. This is how it is currently programmed. I will think of including time instead of number of twilights.

 

 

  • In the plot of the locations (map.FLightR.ggmap(Result)), it doesn't look like all locations are plotted (I have 521 coordinates in the coords.csv file, but the map seems to only show around 70 points). What locations are plotted?

Function plots medians and they overlay (because of the grid). Another potential source of problems is automated zooming. Try zooming out to check whether there is not points outside the plotted area (this should not happen but just in case..)

 

 

  • Which coordinates from the coords.csv file are plotted (Medianlat/Medianlons or Meanlat/Meanlons)?

 

  There is no difference.

 

  • In the plots of lat and long, what do the dark grey and pale grey limits show (are these first and third percentiles respectively?)

  These are 95% and 50% credible intervals.

 

  Cheers,

    Eldar

PS - You better ask short questions - it is hard to find time for such a long reply :)

 

 

 

 







Also tagged with one or more of these keywords: FlightR, staging areas, map

0 user(s) are reading this topic

0 members(s), 0 guests(s) and 0 anonymous member(s)