With a large volume of semantic data and their fast growth in semantic cities, significant challenges in performing efficient and scalable reasoning has emerged in diverse domains. When dealing with large-scale ontologies, the performance of traditional centralized reasoning methods is not sufficient, distributed reasoning methods have thus emerged to improve the scalability and efficiency of inferences. In this paper, an incremental and distributed reasoning method for large-scale ontologies is proposed to realize high-performance reasoning and online query. A novel representation method, transfer reasoning tree and underived assertional triples, is presented to store the incremental ontologies more efficiently, based on which the reasoning process is accelerated and ontology inconsistency is recovered. Finally, a system is implemented on Hadoop and Amazon Cloud, and its application in healthcare validates the effectiveness of the proposed approach.