Two distinct approaches for updating a CMYK printer model in response to systematic changes in print device behavior are presented. In the first method, a corrective model is constructed from a sparse set of newly acquired characterization data and used in addition to the initial printer model. A number of corrective models are investigated, including linear, quadratic, and artificial neural network models. The second method involves directly updating the parameters within the printer model. The updated model parameters are obtained using both the original characterization data and a set of newly acquired data. Both methods are evaluated in a set of experiments in which either the paper stock or the cyan toner cartridge is changed. The corrective model approach is found to be the most effective. The most successful corrective models removed between 76% and 100% of the systematic error.