I have a ~25 year dataset of annual count of species abundance in several traps. Within a year, the abundance in a trap might be dependent on the abundance in another trap.

I want to test the effect of several variables on the species abundance, so I ran a GLM with "Trap" and "Year" as random effects. I suspect that these variables influence the spatial structure of the abundance and that they will detrend my data. But I would like to check if there is still spatial autocorrelation in my residuals.

So I thought about two strategies :

Running a Moran's I on my residuals for each year. The weight of the connections between my points is the inverse of the distance between them. If i'm not wrong it could help me answer the question "Will the number of total individuals be me more alike in close traps than in distant traps" ?

Running a Mantel test for each year, evaluating the correlation between a matrix of geographical distance and a matrix of dissimilarities (that means, the difference in the residuals for each pair of trap). This could help answer the question "Will close traps have more similar abundances " ?

I may be wrong about what these tests do and please correct me if I am. But I would have two questions :

I'm still confused about the difference between the conclusions of these two tests. I'm sure they don't answer the same question but I can't get the nuance (and I don't know which one I should use here). Would someone have a nice example to explain the difference ?

So let's say I use one of those two tests, and that I have only non-significant results for every year except one (so : a spatial structure is detected for only one year). What should I conclude ? Is there a way to know if that single-year autocorrelation is significant "globally" ? Should I modify my initial model ?