Hutchinson definition of the niche has had a tremendous impact on niche modelling and inspired the most prolific species distribution models (SDMs) ever. Those models, called correlative SDM, relate statistically the species presence/absence to various environmental descriptors. They are very much specific and precise, but lack some generality and realisms in the sense of Levins. In this talk I propose another definition of the niche based on species traits and present a new generation of SDMs, namely process-based SMDs, that I use to predict the geographical distribution of forest tree species in Europe. Process-based models are sometimes less precise than correlative models, but are more realistic and more general. I will present some recent results on projections of species future distribution changes as well as projections of past changes (Holocene). Finally I will present how such models can be used to study the impact of phenotypic plasticity on range size and distribution changes, as well as current and future selection gradient of key adaptive traits.