When I modelled the presence–absence of one species as a binomial GAM with a logistic link function I did notice in the plots two discrete packets of points that behave differently (Figure). The smooth functions were constructed as cubic splines and their optimal shape was estimated by minimizing the general cross validation (GCV) criterion. I also modified the analysis with the smoothing method REML (mixed model) with no changes. There is clearly something going on with my data that needs to be talked about. Can you help me to understand a possible explanation for this two clouds of points?