In view of the scattered information resources, highly professional knowledge, and the unreliability and uncertainty in gaining network meteorological information accurately, a meteorological information retrieval system is designed based on semantics, which provides a human computer interaction platform for users. The system mainly uses the semantic retrieval and semantic extension capabilities based on ontology. Through building the meteorological knowledge ontology, an ontology base of meteorological information is set, and by using the inference rules, the explaining properties of the retrieved entry, such as description, keywords, etc., can be queried, as well as the extended queries about its brother, father, and subclass entries. By constantly adjust search terms in the process of interaction, the accurate obtaining of meteorological knowledge is realized. The experimental results show that the system can effectively improve the utilization rate of meteorological hydrological information and improve the accuracy of access to information for users.