Skilful predictions of multi-year US hurricane insured losses by decadal prediction systems

North Atlantic tropical cyclones are the costliest natural hazard affecting the US, and are capable of causing hundreds of billions of dollars of insured losses in a single season.  Tropical cyclone activity has been observed to show considerable decadal variability, linked with variations in sea surface temperatures in regions of the North Atlantic such as the main hurricane development region (MDR) and sub-polar gyre (SPG).

In this presentation we show that a multi-model ensemble of decadal prediction systems can skilfully predict north Atlantic hurricane activity and consequent US insured losses on multi-annual timescales, with a correlation coefficient of greater than 0.7 for 5 year mean hurricane activity.  Rather than tracking tropical cyclones directly in the dynamical models, we make predictions using an index based on predicted temperatures over the north Atlantic.  The skill of the dynamical models outperforms persistence, and could aid decision making for the (re)insurance industry over the US.  As part of the Copernicus Climate Change Service, a publicly available probabilistic forecast of 5 year mean north Atlantic hurricane activity and US insured losses has been produced and will be presented here.