Everyday massive amounts of geo-tagged information are generated around urban environment using micro-blogging services and content sharing platforms. These new Big Geospatial Data sources provide an opportunity to understand people activities and their interaction with the urban environment. In this regard, it is crucial to integrate geo-tagged micro-data with more authoritative sources such as official landuse maps. This integration would benefit the urban research community by combining real time information about people activities and their spatial interaction with the synoptic view of physical infrastructure as depicted in official landuse maps. However, the scientific effort for integrating heterogeneous data sources is hindered by the lack of scalable Geospatial synthesis capabilities to accommodate the massive volume and fast update of microdata. We developed UrbanFlow, a platform to integrate geolocated Twitter data with detailed landuse map (parcel level) to detect and analyze individual human mobility patterns. The platform provides scientists with a set of tools to extract key locations of each Twitter user, assess the extraction quality and uncertainty, and analyze city neighbors' connectivity based on detected users' frequent visitation patterns. These capabilities are built on a novel scalable solution for the point in/nearest polygon algorithm, implemented on Hadoop to harness the power of distributed systems to combine massive point data and large number of polygon in scale-out fashion. Our results showed that we are able to effectively process large data stream of Twitter data (2.42 billion tweets) and synthesize that with highly detailed landuse map (468,641 parcels for the city of Chicago).