This paper presents a multi-resolution approach for tumor motion modeling as a function of respiratory motion of the patient for the purpose of conformal radiation therapy. Respiration induced tumor motion can distort the shape of the tumor, degrade the anatomic position reproducibility during imaging, and necessitate larger margins during radiation therapy planning which may be harmful for healthy tissue surrounding the tumor. The key idea of our approach is the powerful averaging process which allows one to blend independent and arbitrary local models to obtain a global model without introducing the discontinuity on the boundaries. These local models are defined independently to each other by the use of classical basis functions like RBF, Fourier Series, Polynomials, Wavelets etc. based upon a-priori information that we may have about local characteristics of the given input-output data. The proposed approach is validated by using experimental data from a porcine lung.