modelling strategy question

Hi all, I have the following issue. Maybe someone has a helpful hint? Data of a human cross over study (dense sampling) on the absolute bioavailability are to be modeled using NONMEM. In one period the drug is administered i.v., in the other s.c.It turns out that after the s.c. administration the kinetics follwo a 1-cmt model, while after the i.v. administration a 2-cmt model is observable (in each case quite clearly). Can anybody recommend how to fit that situation simulatneously? Thanks in advance, Emily