The purpose of this study is to compare the lung tumor characteristics on Four-Dimensional Cone Beam Computed Tomography (4DCBCT), used for daily patient localization, with Four-Dimensional Computed Tomography (4DCT) used in treatment planning. A programmable motion phantom was used to simulate patients' respiratory motion and loalization based on both 4DCT and 4DCBCT. Two motion patterns, one a regular sinusoidal motion and the other an irregular pattern, were included in the study. The regular motion pattern represented a perfectly sinusoidal respiratory motion with amplitude of 14.5 mm. The irregular motion pattern, which was collected from a real patient, had average target motion amplitude of 15.2 mm with the minimum and maximum motion amplitude of 6.7 mm and 21.5 mm respectively. Our results for regular sinusoidal motion suggest that the 4DCT and 4DCBCT can be appropriately aligned on each individual respiratory phase or on MIP images. The error of alignment is less than 1mm. For irregular respiratory motion, both modalities underestimate tumor motion. In this case 4DCBCT provided a better representation of temporal target shape than 4DCT. There was a difference in location of the target centroid motion between both modalities due to image distortion cause by irregular motion (1.4 mm and 2.8 mm for block and cylinder target objects respectively). We concluded that 4DCT and 4DCBCT can be used to provide good patient alignment with regular breathing. However, the use of MIP and/or end exhalation phase images improves localization in those patients exhibiting irregular breathing pattern. Additionally, methods which can improve regularity of breathing (e.g., audio or visual coaching) should be considered for these patients.