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

plot_slopes_by_location


Recommended Posts

Posted

Hi all,

 I have a few geolocator files that I have not been able to process ( The location estimates are scattered all across the specified grid) and I believe it has something to do with the plot_slopes_by_location function. I know that this function is just to determine whenever the bird leaves the study site, however, in the graph it revealed that there are no dusk estimates after September (see attached picture). All of the geolocators that I have been unable to analyze have this similar problem. I'm not sure if that's actually the root of my problems or if its something else, this is just my best guess. Is anybody aware if this could be the problem and if it is is there anyway to correct for this problem?

Thanks,

Devin

slope.JPG

Posted

Hi Mendez,

how did you call dawn and dusk times before loading the data into FlightR. Have you used the BAStag package? If yes, were the missing dusk events correctly selected by the package for these tags?

Did you possibly subset the data before using plot_slopes_by_location? Maybe something went wrong there?

All the best and have a nice day!

Hendrik

 

Posted

  Hi Devin,

 Absence of the slopes after start of migration is normal, so no worries here. But what I see in the graph is difference between slopes during dusk and dawn yo have in the summer. THis could happen because (1) the start coordianates are not correct, (2) bird moved to a new place (and thus you do not know where to calibrate you tag), (3) there is a specific bird behaviour, that makes this difference. 

 I would suggest you double check the calibration location and if it was correct try try with calibration extracted from a tag that worked. For Intigeo tags that works well.

     Eldar

Posted

Hi Eldar,

I double checked the start coordinates and they are correct. I went ahead and used the calibration from a tag that worked and I was able to get consistent results. The geos that I was having trouble with were attached to females. I was thinking that since the females spend the majority of their time on their nests during the times that I was using for the calibration period that it may cause extreme shading resulting in the unit believing that it was in another location leading to the calibration of the unit being wrong. Is this a viable hypothesis for these birds?

Thanks again,

Devin

Posted

  Hi Devin,

 Yes, this happens often. In many cases (e.g. swifts) calibration twilights could be completely missing due to behaviour.

  Eldar

Posted

I am still having some difficulties analyzing some of the geolocators that have differences between slopes in dawn and dusk. I have tried using the find.stationary.location function, but it provides me with a location that is way different than the original coordinates. All of the geos i'm having difficulties with are females from 2012-2013 and i'm imagining the problem is a lot of shading events on the breeding site which I have been using for the calibration periods.  Even when plotting the slopes with the derived location from the find. stationary.location there is still a lot of differences between the slopes. In the tool help it says that it does not work well with shaded data which i believe is the case.

Eldar recommended above to use the calibration extracted from a tag that worked. Should I try using a calibration file that was extracted from the same year? Also while using the calibration file from another geo i'm still not confident that i'm doing it correctly. 

I took the proc.data from a working geo and calibrated it normally. I would then plug in the proc.data from the geo that did not possess a calibration period into make.prerun.object with the calibration file from the working geo. Is this the correct workflow? I used this method with one of the problem geos and it appeared to work, but upon closer observation the dates did not align at all with the units previously analyzed.

Posted

  Hi Devin,

 Plug calibration from the working tag into 

all.in<-make.prerun.object(proc.data, Grid, Calibration=Calibration)

  Hope this helps,

   Eldar

  • 1 month later...
Posted

Hi all,

Following Eldars advice above I was able to get the other units to work using the calibration of another geo (Thanks again Eldar!). I'm still having difficulties with this one unit and i'm not really sure why. Whenever I run this unit through FlightR it appears that the bird never returned to the breeding ground in the 2010 season, which is possible since we never relocated or captured it, I just think that it's highly unlikely. My first thought was that this unit contains two years of data so I tried splitting it. Once the unit was converted into TAGS format I tried splitting the data in half to make it similar to a geo that had a single year of data. This seemed to work slightly better than trying to run it as a single unit, but the results were still scattered everywhere. All of these failed attempts I was using a different geo for the calibration, I cant run the calibration otherwise. I'm not sure what else to do, any advice would be greatly appreciated!

Capture.JPG

Capture2.JPG

Posted

  Hi Devin,

This does not look good at all, sorry. What happens if you run the GeoLight on this tag? Do you get any stationary period? If so I would suggest you take start and end time from this period and use them to find potential stationary location. After that you can try use this location for calibration. 

  Hope this helps at least a bit.

   Eldar

  • 2 years later...
Posted

Hi all,

My concern is related to posts on “plot_slopes_by_location” .

I am currently working on GLS data (old BAS tags) on seabirds (albatross) deployed for 2-4 yrs and some of them are circumpolar navigating during the non-breeding period.

 

1) The first issue is that I do not have properly calibration data, then I tried to use approximate calibration from a very short period, different loggers et site ~ 80km from the deployment site.

I see that you suggest to use the calibration extracted from a tag that worked, but I failed to understand how you do that.

2) using my approximate calibration data, I process as follow:

                start=c(77.53, -38.71)

                log.light.borders=c(1.5, 9) # default values for Intigeo tag

                log.irrad.borders=c(-3, 3) # default values for Intigeo tag

Calibration.periods<-data.frame(calibration.start=as.POSIXct("2010-01-01"),

                                calibration.stop=as.POSIXct("2020-01-01"),

                                                                                                                                lon=start[1], lat=start[2])

calibration.parameters<-get.calibration.parameters(Calibration.periods, Proc.data,

                            model.ageing=F, log.light.borders=log.light.borders,

                                                                                                                log.irrad.borders=log.irrad.borders)

 

then I tried both make.calibration options without any changes in the results:

Calibration<-make.calibration(Proc.data, Calibration.periods, likelihood.correction="auto")

Calibration<-make.calibration(Proc.data, Calibration.periods, likelihood.correction=FALSE, model.ageing=TRUE, plot.final = T)

 

As you can see in the attached plot, it failed to estimate slope in early 2014 (corresponding to the period where the bird cross the change of datetime line).

Thus I run the end of the script:

all.in<-make.prerun.object(FLightR.data, Grid, start=c(77.53, -38.71),

                                                Calibration=Calibration,threads=1, M.mean=750)                # reduce to one core=node

nParticles=1e4 # for test ~5-6h pour n=2529

Result<-run.particle.filter(all.in, threads=1,

            nParticles=nParticles, known.last=TRUE,

                precision.sd=25, check.outliers=F,

            b=1700)

But it resulted on erroneous location estimates for the ending part of the track.

I try also to cut my original file in before/after files based on the crossing of datetime line without any difference.

 

Thanks

Best

Karine

dawn-dusk_BS29555-23550_flightr_trial2.png

dawn-dusk_BS29555-23550_all_flightr.png

Posted
On 11/13/2020 at 10:35 AM, karde said:

Hi all,

My concern is related to posts on “plot_slopes_by_location” .

I am currently working on GLS data (old BAS tags) on seabirds (albatross) deployed for 2-4 yrs and some of them are circumpolar navigating during the non-breeding period.

 

1) The first issue is that I do not have properly calibration data, then I tried to use approximate calibration from a very short period, different loggers et site ~ 80km from the deployment site.

I see that you suggest to use the calibration extracted from a tag that worked, but I failed to understand how you do that.

Hi, 

before calling make.prerun.object you should load saved calibration file for a tag that worked (load(file= path to *Calibration.RData)

 

 

On 11/13/2020 at 10:35 AM, karde said:

2) using my approximate calibration data, I process as follow:

                start=c(77.53, -38.71)

                log.light.borders=c(1.5, 9) # default values for Intigeo tag

                log.irrad.borders=c(-3, 3) # default values for Intigeo tag

Calibration.periods<-data.frame(calibration.start=as.POSIXct("2010-01-01"),

                                calibration.stop=as.POSIXct("2020-01-01"),

                                                                                                                                lon=start[1], lat=start[2])

I think your calibration period is not properly set. Now it is 10 years used for calibration... You should use here only periods with known location of tag.

Archived

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

×
×
  • Create New...