From what I know the sum to zero constraint is there to avoid the linear dependence between the intercept term and the spatial random effects… However, if I remove it:
