This article engages with debates about the status and geographies of colonial science by arguing for the significance of meteorological knowledge making in late nineteenth- and early twentieth-century Mauritius. The article focuses on how tropical storms were imagined, theorized and anticipated by an isolated - but by no means peripheral - cast of meteorologists who positioned Mauritius as an important centre of calculation in an expanding infrastructure of maritime meteorology. Charles Meldrum in particular earned renown in the mid-nineteenth century for theoretical insights into cyclone behaviour and for achieving an unprecedented spatial reach in synoptic meteorology. But as the influx of weather data dried up towards the end of the century, attention turned to developing practices of 'single-station forecasting', by which cyclones might be foreseen and predicted not through extended observational networks, but by careful study of the behaviour of one set of instruments in one place. These practices created new moral economies of risk and responsibility, as well as a 'poetry', as one meteorologist described it, in the instrumental, sensory and imaginative engagement with a violent atmospheric environment. Colonial Indian Ocean 'cyclonology' offers an opportunity to reflect on how the physical, economic and cultural geographies of an island colony combined to produce spaces of weather observation defined by both connection and disconnection, the latter to be overcome not only by infrastructure, but also by the imagination.