By Laura Tomkins, Senior Catastrophe Research Analyst, Willis Research Network
For the insurance industry, the most telling statistic was a different one: fewer than one in five inundated residential buildings had flood insurance (NYC, 2022). The financial burden fell on the National Flood Insurance Program (NFIP), on federal disaster funds, and on hundreds of thousands of individuals and businesses who had no coverage at all.
A storm that exposed more than coastline
Most of the damage from Sandy was not wind-related as the storm had weakened to a post-tropical cyclone by the time it made landfall. Storm surge was the driving impact from Hurricane Sandy. Several factors converged in a particularly damaging way: the storm's unusually large wind field, its near-perpendicular track toward the coast, the funnel-like geometry of New York Harbour, and near-coincident timing with high tide. The result was a surge reaching up to 4.3 meters in some locations, inundating areas outside of Federal Emergency Management Agency (FEMA)-mapped flood zones ( ProPublica 2013). Sandy’s flooding was not a failure of meteorological forecasting; the track and timing of Sandy were well-predicted days in advance. It was a failure of risk translation. Atmospheric science could describe what had happened, but the insurance industry could not adequately estimate the chances of it happening again.
Flood risk was different — and the market knew it
Flood had long been the uncomfortable outlier in catastrophe modelling. Wind risk, the dominant focus of the catastrophe modelling industry since Hurricane Andrew in 1992, can be characterized at regional scales with reasonable confidence while flood cannot. It is fundamentally local: a function of precise elevation, drainage, soil saturation, and coastal geometry at the scale of individual streets and properties. Capturing these features required high-resolution hydraulic models that, through most of the 2000s, were simply too computationally expensive to run at the scales the insurance market needed. In many of the areas worst affected by Sandy, flood maps had not been updated since the 1980s; created with technology that could not capture risk at the scale of individual properties ( ProPublica 2013). As a result, flood risk was either excluded from private coverage, crudely approximated, or offloaded to government-backed insurance programs like the NFIP. The tools necessary to improve flood modeling existed in academic hydrology, but they were slow, data-hungry, and built for individual river catchments rather than the kind of large-scale risk assessment that the (re)insurance industry needed. What the market needed was a flood model that could work at scale. In 2012, that tool was closer than most in the industry realized.
WRN-funded research had been working on exactly this problem
Since the late 1990s, Professor Paul Bates and his team at the University of Bristol had been developing a flood inundation model called LISFLOOD-FP. The core insight was counterintuitive. Deliberately simplifying the underlying physics produced a model that was not only faster and cheaper to run but could be applied at continental and even global scales without sacrificing practical accuracy (Bates et al., 2010). The prevailing assumption had been that better flood models required more complex physics and more computing power; however, Bates showed that a simpler framework could do more with less.
With support from WRN, this work was developed and validated across flood risk settings in Europe and Southeast Asia (U. Bristol, 2012). By 2012, it had already been adopted commercially: JBA Consulting had developed LISFLOOD-FP-based models to build national flood risk maps for the UK Environment Agency (U. Bristol, 2012). That same year, the research was awarded the Lloyd's of London Science of Risk Prize for Natural Hazards, recognition from the insurance market itself that this was science with direct commercial value (Lloyds 2012).
Sandy’s legacy: Science advanced the protection gap did not
More than a decade on, science has progressed considerably. Global high-resolution flood datasets now exist that would have been computationally unthinkable in 2012. Perhaps the clearest demonstration of that progress is how quickly flood science moved from academia into operational decision-making. In 2013, Bates co-founded Fathom alongside colleagues from the University of Bristol, translating the research directly into a commercial venture. Fathom became one of the leading flood analytics platforms used by (re)insurers and climate risk practitioners worldwide, and was subsequently acquired by Swiss Re.
However, the protection gap has not closed; in some respects, it has widened. Flood insurance take-up rates in areas affected by Sandy have declined since 2012, not grown (RMS, 2022).
What 2012 demonstrated, and what remains as true in 2026 as it was then, is that better flood models are a necessary but not sufficient condition for a functioning flood insurance market. The models inform us where the water will go – the harder problem is developing the products, incentives, and public understanding that translate that knowledge into coverage. That is a challenge the WRN and its partners continue to work on, and one that every major coastal flood event since Sandy has made more urgent.
Key takeaways
Beyond the devastating loss of life and property, Sandy illustrated a fundamental gap in flood insurance coverage as fewer than one in five flooded homes were insured, leaving the financial burden on individuals and government programs rather than markets equipped to absorb it
WRN-funded research into high-resolution flood inundation modelling, recognized by the Lloyd's Science of Risk Prize in 2012, had already begun addressing the scientific foundations of that gap
Flood modelling has since matured into a global industry, but the protection gap has not closed, and translating better science into better coverage remains the defining challenge
|
