The capability to forecast pavement crack performance has been identified as a critical pavement management system (PMS) need. In particular, crack performance forecasting can support proactive pavement management strategies. However, previous attempts at developing crack forecasting models have met with limited success. Because of the complexity of the pavement crack progression mechanism over time, a dynamic forecasting model could be more appropriate as opposed to static forecasting models. The results of a research effort to develop an adaptive filter model to forecast a pavement crack index for Florida's highway network are summarized. The concept of the adaptive filter model is introduced along with its mathematical algorithm and modeling procedure. The adaptive filter model, which was developed with field data collected from Florida's highway network, is able to dynamically forecast crack performance on the basis of historical crack performance data. For the purpose of model evaluation, the adaptive filter model was compared with the traditional autoregressive (AR) model. According to the comparison results, the adaptive filter model had significantly lower errors as compared with the AR model, demonstrating better forecasting capability. The adaptive filter model was more responsive to unexpected changes in pavement performance than the AR model. Given these advantages, it appears that such an adaptive filter model may have considerable potential for use in PMS applications.