Integrating updates from timely sources such as volunteered geographic information (VGI) into the spatial data maintained at official agencies is becoming a more demanding requirement but presents many challenges. This paper proposes an approach to addressing the technical challenge of propagating updates from timely sources (e.g. OpenStreetMap) to spatial data maintained at separate map scales. The main idea is to establish a multiple representation database (MRDB) for datasets at different scales and time to facilitate incremental update, where linkages between corresponding objects at different datasets are made explicit. First, two ways in which the timely sources can be integrated into official data for incremental update are discussed. To derive the linkages between different datasets, a data matching procedure based on computer vision is presented and fine-tuned to match data in different scale ranges. Furthermore, the generalization history used to produce smaller scale data from the larger ones in official data is inferred based on the linkages, and is then used to guide the update propagation. Finally, a framework for incremental generalization in MRDBs is proposed, where crucial issues like strategies for update propagation, cartographic generalization, and the so-called ‘chain reaction’ are addressed. The framework is implemented as a fully automated process where operators like simplification, enlargement, compression, displacement and typification are incorporated into the incremental update process. By testing the framework against real world data sets (i.e. OpenStreetMap and official data at 1:10k, 1:50k and 1:100k), we show that the updates are integrated consistently into existing data in terms of spatial relations and cartographic quality. Our work suggests that making use of timely sources by official mapping agencies and companies in a continuous or event-driven data update is technically feasible, with further improvement and extensions discussed.