A major problem in operational land cover change detection using remotely sensed data is to separate the change signals caused by land cover changes from those due to vegetation phenology. This study provides an approach to this problem by systematically analyzing the spectral properties of major land cover change processes and the phenological profiles of different land cover types. The phenological profiles were derived from a global data set consisting of a full phenological year's data record of the 1 km monthly composites from the Advanced Very High Resolution Radiometer (AVHRR), while land cover change signals were simulated from the spectral signatures of corresponding land cover types in different seasons. A decision tree method was used to derive the decision rules that provide best separation between the change signals of land cover changes and vegetation phenology. These decision rules were referred to as land cover change trajectories. A complete set of change trajectories was developed for the globe in all seasons of a phenological year. Results from this study indicate that during most seasons of a phenological year, major land cover change processes including deforestation, denudation, revegetation, flooding, flood receding and vegetation burning, can be separated form one another and from vegetation phenology in the red-near infrared space. The derived trajectories, when validated, can serve as a theoretical basis for developing operational change detection algorithms.