I used Generalized Estimating Equations (geepack in R) on my regularly monitored (twice a year) plant abundance data in grid cells to evaluate their trends of density.

Since these are time-series data, GEE was an ideal non-parametric method to implement, because it takes the pairwise correlations between time points (correlation structure) in account.

But my data and the residuals of GEE are spatially correlated as well.

I found the spind package for R, which handles spatial auto-correlated data in GEE, but it seems that this function only deals with spatial autocorrelation, and the available correlation structures are very restricted (e.g. cannot choose "ar1" anymore, or use my pre-calculated correlation matrix as a "fixed" ).

So GEE can only handle temporal OR spatial autocorrelation but not BOTH?

Did I get it wrong?

Is there any way to solve this in GEE?

Thank you for any idea!