Incompleteness of natural disaster data and its implications on the interpretation of trends

We use data from a leading global database, primarily the reported numbers of fatal 'hydrological', 'meteorological' and 'geophysical' disasters and associated fatalities, to elucidate long-term trends in natural disaster impact. There are very strong upward trends in the number of reported disasters. However, we show that these trends are strongly biased by progressively improving reporting. We apply well-established methods based on analysing evolving differences in the patterns of large (many fatalities) to small (fewer fatalities) events to assess and quantify this bias. For all disaster types investigated, the long-term (multi-decade) true number of disasters appears to be unexpectedly stable over time. There was, however, a clear temporary increase between 1980 and 1990 for hydrological and meteorological disasters. Importantly, there has been a large long-term decline in total fatalities reported for weather-related disasters, attributable to very effective mitigation for larger disasters. Probably due to improved mitigation rather than reduced hazard, data for the most recent decades shows a systematic decrease in the number of larger (more fatality) weather-related disasters compared to smaller ones. In contrast, for the geophysical disasters, predominantly earthquakes, the relative numbers of larger versus smaller disasters, are fairly constant, as are total fatalities.