Records of prehistoric tropical cyclones occur in the form of ridges of coral rubble, sand, shell, sand and shell, and pumice; erosional terraces in raised gravel beaches; barrier washover deposits; and, sediments deposited in the shallow offshore marine environment. Other less well-documented records occur as variations in isotopic ratios within speleothems and possibly tree rings, and changes in pollen records resulting from introduction of new species after forest disturbance due to cyclonic winds. As yet, such records have not been identified beyond 5500 years of age. Recent palaeotempestological studies in the United States and northern Australia have highlighted that the frequency and magnitude of these natural hazards do not remain constant over time, and there are periods when cyclogenesis is enhanced, and others of relative quiescence. Recognition of such regime changes, or non-stationarity in the long-term record, is important for risk assessments of this hazard. Until now however, few if any risk assessments have incorporated data from the long-term record of tropical cyclones, and instead have relied on generally short instrumented historical records. The longer-term records suggest that such an approach may miscalculate the 1% Annual Exceedance Probability risk to coastal communities from future tropical cyclones.