Hi Søren,
Your second best model (M4) has a fairly good AIC (~3 units above the best
model) and has both of your predictors. If one of your study’s goal is to
quantify the effect of both of your predictors, then it’s a good reason to use
your model M4. Having them both together is important if
Thanks for the comment Cecile
You are completely correct that Climbing and Rainforest is correlated which
explains the "odd" behavior when both are added to the model.
It may just be a case of Walds test giving a much worse approximation than the
AIC, but I was just surprised in this case