Social networks today have emerged as hotbeds of online user conversations. Social microblog sites like Twitter have become favorite portals for users to discuss and express opinions on events and topics. Established event detection techniques on microblog streams today are capable of detecting events early in their lifecycle, amidst the volumes of user message exchanges. Techniques have been proposed in literature to identify topical conversations on microblogging portals comprising of unstructured data with no explicit discussion thread, distinguishing such conversations from isolated expressions of topical interests. However, evolutions of discussion topics have not been studied in a geographical context before. In the current work, we identify and characterize topical discussions at different geographical granularities, such as countries and cities. We observe geographical localization of evolution of topical discussions. Experimental results suggest that these discussion threads tend to evolve more strongly over geographically finer granularities: they evolve more at city levels compared to country levels, and more at country levels compared to globally. Our algorithm to find geographical evolution of discussion sequences and the derived insights can be used for information spread analyses and related applications on microblogging networks.