Natural disasters trigger complex chains of events within human societies¹. Immediate deaths and damage are directly observed after a disaster and are widely studied, but delayed downstream outcomes, indirectly caused by the disaster, are difficult to trace back to the initial event^1,2. Tropical cyclones (TCs)—that is, hurricanes and tropical storms—are widespread globally and have lasting economic impacts^3,4,5, but their full health impact remains unknown. Here we conduct a large-scale evaluation of long-term effects of TCs on human mortality in the contiguous United States (CONUS) for all TCs between 1930 and 2015. We observe a robust increase in excess mortality that persists for 15 years after each geophysical event. We estimate that the average TC generates 7,000–11,000 excess deaths, exceeding the average of 24 immediate deaths reported in government statistics^6,7. Tracking the effects of 501 historical storms, we compute that the TC climate of CONUS imposes an undocumented mortality burden that explains a substantial fraction of the higher mortality rates along the Atlantic coast and is equal to roughly 3.2–5.1% of all deaths. These findings suggest that the TC climate, previously thought to be unimportant for broader public health outcomes, is a meaningful underlying driver for the distribution of mortality risk in CONUS, especially among infants (less than 1 year of age), people 1–44 years of age, and the Black population. Understanding why TCs induce this excess mortality is likely to yield substantial health benefits.