Hi everyone,
I'm applying a dynamic multi-season occupancy model to a rut dwelling toad species. I have three primary surveys (in 2018, 2020 and 2022), each with three secondary surveys. As it happens, 2020 and 2022 have been very dry years over here, with some extreme droughts in some areas, and since many ruts turned out to have dried out (especially in 2020), the occupancy has drastically dropped in 2020, and is still rather low this year. Ruts are my sampling unit.
To estimate Gamma (col) and Epsilon (ext), I incorporated a dummy variable that I called "sec_yearly", where 1 means that the rut has been dry at all three secondary surveys within a year, and 0 means that there was water in the rut at least once within a year. Here are the estimates I found:
Backtransformed linear combination(s) of Colonization estimate(s)
Estimate SE LinComb (Intercept) Sec_yearly1
0.329 0.0469 -0.712 1 0
Backtransformed linear combination(s) of Colonization estimate(s)
Estimate SE LinComb (Intercept) Sec_yearly1
0.496 0.262 -0.0149 1 1
Now, I think I might misunderstand something. From this output, I'd read something like "the colonisation prob is ~0.3 when sec_yearly=0, and ~0.5 when sec_yearly=1, hence colonisation is more probable when a rut is dry". Extinction output gives something similarly counter-intuitive.
Obviously this cannot be right.
Would it be more correct to understand that as one year shows a lot of dried ruts, and consequently a lot of absence in my dataset, chances of a colonisation event are higher at the next survey?
I have a hard time picturing if this makes any sense from a biological point of view. If it's the correct way to interpret these parameters, it seems to me they're very dependent from meteorological conditions differential between years. Maybe an interaction term (sec_yearly*rainfall for exemple) would be more appropriate.
This is more of a generalist question about how to interpret dynamic parameters I guess, but should you need more details about the model I'm building I'd be happy to provide them.
Thank you,
Vincent