We propose a time-series model for characterizing land cover change. The model is developed based on a realistic view of land surface transformation - spatially discrete land cover change events are continuous processes over time. The model utilizes a non-linear mathematical function to reconstruct the underlying continuous land change process from temporally discrete satellite observations. Based on the coefficients of the mathematical function, the model simultaneously derives the timing, intensity and duration of land cover change. Moreover, the smoothing feature of the model can also be used to remove random noise in time-series land cover data and hence to re-establish the stable land cover states before and after change. We illustrate the application of the model by showing examples of characterizing deforestation using time-series MODIS data as well as characterizing urban land expansion using time-series Landsat data.