variogram with error = TRUE returns empty plot

[Allie Anderson]

Hi Chris and team,

I'm wondering if you could help me sort out an issue I've been having creating variograms while accounting for spatial error. The main issue is when I try to add error = TRUE into the variogram() function I receive a largely empty plot, whereas when I run without error (default), the variogram plots as normal and makes sense. 

The data I'm running through are Argos data (and some individuals where tags collected both Argos and GPS data). Data are points that occur during a stopover period for individual birds. I notice normal variograms with the default but for most stopover events, I receive empty plots. 

Here's an example for a single bird. It has 336 points that occur over a period of 8 days. The data are newer Argos with the semi major and semi minor ellipse info (I do also have birds that only have LC class codes)

The tlm summary looks like this:
$identity

[1] "STOP_1" $timezone [1] "UTC" $projection [1] "+proj=tpeqd +lat_1=69.4293123249529 +lon_1=-98.2208819681596 +lat_2=53.1062998104261 +lon_2=-82.0369529047841 +x_0=0 +y_0=0 +datum=WGS84 +units=m +no_defs" $`sampling interval (minutes)` [1] 33.48333 $`sampling period (days)` [1] 8.94625 $`longitude range` [1] -108.1709 -106.5596 $`latitude range` [1] 71.47301 72.04295

I can see the ellipses when I plot the data using:
plot(example_tlm, col = "blue", trans = 0) 

It looks like this. Some of the points have very large ellipses on the low quality classes:

When I create the variogram without error 

vario_noerror <- variogram(example_tlm)
plot(vario_noerror)

When I plot with error = TRUE using:
vario_error <- variogram(example_tlm, error = TRUE)
plot(vario_error)

Here's the output. Any ideas about what's happening in this case? It looks like there are a couple of fast speed points that could maybe be removed as outliers. 

Any tips greatly appreciated!
Allie

[Allie Anderson]

Hi Chris et al,

I dove into this a bit more to explore what's happening. I think this is an issue of large error ellipses on some of the observations. When I remove outliers using outlie(), the same pattern arises, so it doesn't appear to be an a speed or distance outlier issue. I ran the same bird through with the Argos LC codes only (removed the ellipse data), and I was able to create variograms with error = TRUE. I also tried a variety of options to filter out observations where ellipses are quite large. One that provided somewhat reasonable output in the variograms with spatial error was filtering out rows where the Argos.semi.major and Argos.semi.minor axes were outliers (identified with the IQR proximity rule on the whole dataset of birds). Does this approach seem reasonable to you, or is there an alternative you might recommend in this situation? Or maybe something else I am missing?

Thanks so much,

Allie