The modernization of urban scenarios includes improvements in wireless communication and efficient content dissemination to citizens. In these scenarios, the demographic densification and the phenomenon of popularization of mobile devices may characterize massive demands for online content in crowded regions, able to temporarily deplete the resources of the network infrastructure. To address this challenge, the next generation of wireless networks envision local cooperation among users using the D2D paradigm to improve content retrieval. Much effort has been dedicated to proposing mechanisms for cooperation and resource sharing; however, the advantages of big data have been little explored to support D2D to cope with urban and social dynamics. In this article, we investigate human mobility using data from online social networks, observing the mobility of 362,000 users during one year in New York City. We analyze the spatiotemporal features of the city, and their effects on encounters between users and content dissemination through D2D. Thus, we propose a framework for distributed caching based on social and spatiotemporal factors. The results of the experiments demonstrate the feasibility of D2D to offload the network demand and performance gain in the provision of content opportunistically.