The temporal dynamics of influenza are driven by climatic conditions and by antigenic evolution of the virus. However, important spatial variation remains unexplained. I will discuss recent work that uses fine-grained insurance claims data to study differences in influenza dynamics across 603 cities in the United States. The data reveal predictable spatial variation in epidemic intensity – some cities consistently have more intense epidemics than others, year after year, including cities within the same climate and antigenic regimes. In particular, we find that incidence in smaller cities is consistently focused on shorter periods of the influenza season, while in larger cities, incidence is consistently more diffuse. We fitted a climate- and antigenically- forced compartmental epidemic model (SEIR model; a low-dimensional system of ODEs) to the incidence data in each city, which revealed that base transmission potential (transmission rate that is not modulated by climate) is positively correlated with population size and with spatiotemporal organization in population density. Counterintuitively, elevated base transmission rate produces a milder response to climate forcing in metropolises, by altering the spatiotemporal geometry of herd immunity.