Variogram interpretation help (please!)

[Jamie Bolam]

Hello all,

This is my first time using CTMM and while I generally understand things with the aid of Calebrese et al 2016's examples and can generally identify which of my collared animals have settled or not, and when, some of the patterns on these variograms have really rumbled me and I would appreciate your clarification! The labels on each photo show different individual.

What does the semi-variance y-axis on the variograms even signify? I have been a bit confused.

- B appears to have explored quite quickly before resuming roughly range residency with a rough but squiggly asymptote. Am I correct in thinking that this means the area used has increased post-collaring, and then is increasing/decreasing over time? Or does it show the individual moving closer to the area of collaring when the semi-variance decreases? What would the decline in semi-variance at the 5 month mark signify? Also, does anyone know why the black line is quite thick with oscillations? And does the lighter grey signify error margins?

- C has a much shorter dataset; what could make the 4 data points at the end go flat?

-E has really confused me as everything looks to be expected until just before 7 months  when the semi-variance hits 0 for a time and then spikes straight up. I presume this is an error of some sort but when I checked the data, nothing looks out of the ordinary (the collar still shows movement etc). Any ideas?

- If I am interpreting K correctly, there was rapid increase in range post-collaring up to range residency, following by a bit of wobble in range size and then either a decrease with the semi-variance going down, or returning closer to where the collar was fitted, and staying more range resident there. Would a lower semi-variance indicate a smaller home range?

- Mu shows expected increase in range size but takes a while to do so. What might the little oscillations mean, and any thoughts on why it might have flatlined (at a higher semi-variance too)? The shape of the curve at the start (convex?) is the opposite to that of K and Mwa - what might this signify?

-Interpreting Mwa's variogram should be fine based on answers to the above

- S's variogram is also very intruiguing and any ideas would be appreciated!

All the best and many thanks,

Jamie

[Christen Fleming]

Hi Jamie,

The semi-variance (y-axis) is proportionately the expected (average) square distance between any two locations with a given time lag (time difference) between them. The time lag (x-axis) is the amount of time between the sampled locations. This is a relative amount of time and not an absolute time (timestamp).

The light gray are the confidence bands of the empirical variogram. These require CI="Gauss" for accuracy beyond the asymptote, which is slow to calculate on large datasets. The default is fast to calculate, but tends to grow a little too wide after the asymptote.

B, Mwa look fairly range resident. The small oscillations may either be some switching in the sampling interval, which can be fixed with the dt argument, or something like a daily periodicity.

C doesn't look resident, but there may be a short segment of settlement in the data.

E looks like the data are mostly resident over a ~6 month period, but contain a migration or dispersal that needs to be segmented.

The spikey behavior at the end is probably from gaps in the data at those lags.

K probably contains a migration or dispersal.

Mu definitely contains a dispersal or migration.

S looks like it contains multiple migrations or dispersals.

In most of these non-resident cases, you can see low-variance stretches of the variogram (like the first half of Mu) and high-variance stretches of the variogram (like the second half of Mu) . The low-variance stretches correspond to the size of the individual home ranges, while the high-variance stretches corresponds to the size of the entire migratory/dispersal range. If you can segment the data appropriately, then the individual variograms of the segments will be flat. Another approach would be to do a sliding-window or seasonal analysis.

Best,

Chris