Location-aware publish/subscribe (pub/sub) has attracted a lot of attentions with the booming of mobile Internet technologies and the rising popularity of smart-phones. Subscribers subscribe their interests with their locations as subscriptions, and publishers publish geo-information as events. Many state-of-art applications with a massive amount of geo-information, such as location-aware targeted advertising systems, face this situation. Existing related work mainly focuses on unstructured geo-textual information. However, many online-to-offline applications have enormous geo-information with different structured descriptions. To handle such structured information, a new type of location-aware pub/sub approach is needed. In this paper, we handle these subscriptions using boolean expressions. Since the number of publishers and subscribers can be enormous, it is extremely important to improve the matching effectiveness and efficiency of top-k query processing. In this paper, we develop a novel solution named RRt-trees. RRt-trees integrates Rt-tree and a predicate index structure together to return top-k best matched subscriptions from a great number of events. Our experimental results on synthetic and real-world datasets show that RRt-trees achieve better performance than baseline methods.