This paper presents the development of probabilistic models for the service life prediction of low-slope roof components (i.e. membrane and flashing). Data used for model development are collected within the Building Envelope Life Cycle Asset Management (BELCAM) research project. The parameters that significantly affect the deterioration of built-up roof (BUR) membrane and flashing are identified through statistical tests (i.e. correlation analysis and ANOVA). Transition probability matrices of Markov-chain deterioration models are developed to be used in predicting the future condition of roof components. Copyright ASCE 2006.