A new approach in loss estimation for correlated hazards
As global temperatures rise, hurricanes are bringing heavier rainfall and causing greater potential for destructive flooding. Understanding how wind, surge, and inland flood interact is essential for loss estimation and financial and societal resilience. Here, Dr David Cross, Principal Analyst and catastrophe model specialist at JBA Risk Management, explains how JBA and Applied Research Associates (ARA) have combined their world-leading models to offer a unified approach to hurricane and flood risk.
The most powerful models are built not just on the best data and science, but on strategic collaboration and shared values. This belief has brought together JBA and Applied Research Associates (ARA) to align their world-leading expertise in flood and hurricane modelling, offering clients a flexible, joined-up framework for hurricane wind, storm surge, and inland flood risk across the United States.
Our approach allows users to draw on the best-in-class modelling they already know and trust – HurLoss® from ARA for hurricane wind and surge and JBA’s inland flood model – while enabling deeper insight when used together. By integrating our strengths, we give insurers, reinsurers, and risk managers choice, coherence, and confidence in understanding the full spectrum of hurricane-related losses.
Why joint modelling matters
Over the last 10 years, flooding from tropical storm rainfall was the single deadliest hazard.
National Oceanic and Atmospheric Administration 2023
Hurricane impacts are growing with climate change (Figure 1). Rising ocean temperatures provide more energy for hurricanes as they spin-up at sea, while rising air temperatures mean more moisture, increasing the potential for extreme rainfall and flooding. Events such as the Texas flooding of July 2025 from Tropical Storm Barry show how tropical moisture brought inland can lead to catastrophic impacts.
These “compound” events – where wind, surge, and inland flood coincide – present a growing challenge for insurers and risk managers. Flooding has always accompanied hurricanes, but a trend toward slower-moving, rain-laden storms (Harvey, Florence) makes joint assessment essential for accurate loss estimation.
The challenge of correlated hurricane losses
Recent hurricanes such as Helene and Milton underline the need for a unified perspective. Helene was the deadliest US mainland hurricane since Katrina (2005) causing an estimated $17.5bn in insured losses, while Milton – the fifth-strongest Atlantic hurricane on record – generated around $20bn in insured losses.
Floods and hurricanes are complex hazards that are typically modelled separately to achieve the best results. For many insurers, this is sufficient and aligns with what they expect from model vendors. But modelling each peril in isolation can lead to underestimation or misrepresentation of tail losses when hazards are correlated, since the total loss variance must be estimated post hoc using assumptions about how these hazards interact. This is explained further in Box 1.
Floods and hurricanes are complex hazards; therefore, they are typically modelled independently to achieve the best results.
Box 1: Why combining separate models can understate tail losses
Imagine running two separate catastrophe models – one for hurricane wind and one for inland flood – each with its own independently generated event set. When the results are combined, there’s no guarantee that the flood event with the largest loss aligns with the wind event with the largest loss, even if in practice, both could have been triggered by the same hurricane.
The result is a combined loss distribution that misrepresents compound extremes. Because unrelated events are blended (e.g., moderate wind with moderate flood), the aggregation introduces artificial smoothing, which dampens the upper tail of the distribution. This leads to underestimated loss volatility, reducing the perceived risk of co-occurring wind and flood losses in extreme years.
In the case of Hurricane Harvey, Lloyds research has shown that insured losses were far lower than total economic losses, which were dominated by flood damage. This disparity arose because wind and flood were treated as separate perils—both in insurance coverage and in risk modelling. While wind risk was widely captured within standard homeowners’ and renters’ policies, flood losses fell largely under the National Flood Insurance Program (NFIP), where uptake was low due to affordability issues and limitations in flood mapping. The result was a fragmented view of hurricane losses that underestimated the combined impact of wind and flood, leaving many communities underinsured.
Two leaders, one vision
Our vision is to bring together the best science and data via two very different modelling disciplines – Hurricane and Flood – to support decision-making in insurance and social resilience. This alignment empowers clients to tailor their risk modelling approach – with confidence that each model reflects best practice, and that their combined use enhances, rather than replaces, standalone value.
Our vision is to bring together the best science and data via two very different modelling disciplines – Hurricane and Flood - to support decision‑making in insurance and social resilience.
JBA’s flood models are founded on a detailed understanding of hydrology and hydraulics, advanced statistical methods, and high-resolution spatial data. Underpinned by time-series modelling, they capture the evolving conditions leading up to flood events. This approach is essential for representing the antecedent conditions that strongly influence catchment response. By simulating continuous precipitation and runoff over long periods, the models produce spatially and temporally coherent flood events, enabling robust assessment of river and surface water flood hazard.
A major challenge with flood modelling is to accurately represent the heterogeneity of flood impacts. Flood is a highly localised hazard where two properties near one another can experience vastly different flood risk. JBA’s high resolution 5m flood maps – made with detailed hydrodynamic river and surface water modelling – are unique in capturing this granularity (Figure 2).
ARA’s hurricane modelling is built on a deep understanding of tropical cyclone dynamics, advanced numerical methods, and high-resolution environmental data, structured around an event-based approach that simulates the full lifecycle of storms. Using physically derived track generation and nonlinear wind field modelling, the framework produces consistent representations of wind, precipitation, and storm surge for each simulated event.
This physics-driven methodology allows for coherent hazard footprints and realistic temporal evolution, enabling detailed assessment of hurricane wind, surge, and rainfall ingress into damaged buildings. Figure 3 shows a sample wind field from ARA’s hurricane modelling over Florida.
Event-based modelling for joint hazard assessment
Our joint solution connects ARA’s hurricane rainfall with JBA’s flood hydrology, creating an event-based workflow that models the full hydrological response of US catchments. Antecedent conditions are first established using reanalysis precipitation, excluding contemporaneous storm activity. ARA’s hurricane rainfall is then introduced, with spatial and temporal scaling to drive JBA’s rainfall-runoff models (see Figure 4 for a schematic illustration of how ARA and JBA’s modelling workflows interact). A nested catchment approach, combined with prediction in ungauged basins, delivers dense and consistent coverage across the entire river network – capturing the full hydrological impact of ARA’s hurricane rainfall.
The process produces hurricane-driven flood events that correlate directly with ARA’s wind and surge simulations. These events are combined with JBA’s existing inland flood catalogue, preserving overall event frequencies while capturing realistic compound losses.
Developing a unified view of hurricane wind and flood risk required close collaboration between JBA and ARA. Our first step was to ensure that the two models – each designed for different aspects of the hazard – spoke the same scientific language. Through joint analysis of hurricane footprints and rainfall patterns, we aligned event data and applied extreme rainfall scaling to ensure consistency between ARA’s hurricane events and JBA’s flood simulations. This process allowed us to build an event-based rainfall–runoff workflow capable of translating hurricane rainfall into realistic catchment responses across the US Figure 5 illustrates how hurricane rainfall - translated to JBA’s grid - feeds into catchment response.
We then brought the two event sets together within a consistent stochastic framework. JBA’s data represent the full range of US rainfall-driven floods, including tropical cyclones, while ARA’s focus is on North Atlantic hurricanes and their wind intensities. By applying ARA’s wind thresholds to refine JBA’s tropical cyclone events, we harmonised storm frequency and landfall timing across both models, producing a coherent representation of combined wind and flood hazard.
The result is a unified, flexible modelling suite for the US – JBA for flood, ARA for hurricane wind and surge – that can be used independently or together for a correlated view of risk. This joint framework gives insurers and reinsurers a consistent basis for assessing the full spectrum of hurricane losses, supporting better underwriting, portfolio management, and reinsurance structuring.
Looking forward
Our collaboration provides a blueprint for the future – one where excellence in specialised domains comes together to deliver a holistic, client-focussed view of risk, without compromising scientific rigour or independence. Both JBA’s and ARA’s models run on the OASIS platform for loss estimation and are available through Verisk’s Model Exchange.
For more information about using the JBA’s inland flood model independently or in tandem with ARA’s HurLoss, get in touch!