Using the dynamic response of trees to estimate local wind speeds could provide valuable insights for the insurance industry to assess risks and understand the potential financial impact from wind hazards.
With extreme windstorms becoming more severe and costly, the need for accurate risk management has never been greater. As of November 2024, the U.S. has experienced a series of extreme wind-related events—including 17 severe convective storms, 4 tropical cyclones, and two winter storms, each causing losses exceeding $1 billion.1 Extratropical cyclones in Europe also contribute significantly, with annual insurance losses reaching around $2 billion. Climate change is expected to make storms stronger across Great Britain and Western Europe,2 further driving up costs. These rising losses highlight the urgent need for better risk management strategies to assess wind speeds at individual high-risk locations.
The challenge of urban complexity
But measuring wind in urban and other complex areas is challenging. Urban environments are difficult to assess due to their dense and intricate designs. Wind speeds and directions can change dramatically between buildings. This complexity grows as city infrastructure becomes more sophisticated, making it harder to capture localized wind patterns. Traditional methods, like stationary anemometers and meteorological stations, often fall short in these settings. They struggle to reflect the real wind conditions at street level or near specific structures.
This lack of precision complicates risk assessments and the development of insurance products that truly reflect the risks faced by urban policyholders. High costs and logistical challenges further limit the use of traditional wind measurement methods in cities.3, 4 This makes the need for advanced data collection even more urgent. Accurate wind data is essential for urban planning and construction. It helps predict wind patterns, create more reliable wind hazard maps, and better understand the impacts of rare storm events on both current and future infrastructure.
The need for detailed and localized data
Catastrophe models are also critical tools for predicting losses from extreme wind events, but they bridge existing data gaps by fitting statistical distributions to claims data and are typically built to provide an aggregate view of risk rather than assess risk at a site-specific level. While these models can identify broad trends and highlight top loss-driver locations, they often miss the details of individual sites. This limitation can lead to over- or under-estimating risk at the site level, potentially resulting in inaccurate insurance pricing and skewed portfolio losses.
Therefore, detailed and localized wind data becomes essential. By complementing catastrophe models with more precise data, insurers can improve the accuracy of loss predictions and better tailor their reserves and (re)insurance strategies. WTW's Climate Practice goes beyond the aggregate approach by offering tailored, site-level solutions. We produce probable maximum losses at chosen return periods to determine the most likely loss that could occur at specific sites. These losses can then be calibrated further with additional localized data, adjusting the initial portfolio modelling to reflect a more granular view of risk, particularly at the high-loss driver sites.
Bridging the gap with digital innovation
This is where innovative research, like the "Seeing extreme winds" project, steps in to fill these gaps. Inspired by an initial study at Stanford,5 "Seeing extreme winds" is a collaboration between the WTW Research Network and Loughborough University, through the TECHNGI-CDT Centre of Doctoral Training program. It addresses the challenges of measuring wind in complex urban settings, where traditional methods often fall short.
The project uses video analytics to analyze the natural movements of trees and flags—elements sensitive to wind and found throughout urban areas—to estimate wind speed. This innovative approach creates detailed, localized wind data without the need for costly, specialized equipment.
A pilot study will validate this concept using machine vision software, Flow On The Go,6 and machine learning to analyze video footage for wind-induced movements. This will demonstrate the potential to leverage everyday technology, such as CCTV cameras, to collect advanced environmental data. Following the success of the pilot study, we envision expanding this methodology to broader areas, such as university campuses, by utilizing fixed CCTV cameras for continuous monitoring. This larger scale analysis is expected to deepen our understanding of wind impacts at a micro-scale, providing insights essential for more accurately mapping wind fields around buildings and other structures.
Shaping the future of wind risk management
As extreme weather events become more frequent and severe, evaluating wind speed intensity at individual high-risk locations is increasingly critical. Research projects like "Seeing extreme winds" are bridging the gap between traditional models and the advanced, localized data needed to improve site-level assessments.
By generating detailed, localized wind hazard data and maps, "Seeing extreme winds" offers insight into how insurers might tailor products more closely to the actual risk profiles of property owners. This, in turn, enhances portfolio risk modelling and supports more informed conversations about limits and deductibles with insurers. Such an approach helps create more accurate, equitable, and adaptable wind risk management strategies for our ever-changing urban landscapes. Given the rising scale of windstorm losses globally, this level of precision is essential.
Footnotes
1. NOAA National Centers for Environmental Information (NCEI). U.S. Billion-Dollar Weather and Climate Disasters. (2024).
2. Priestley, M. D. K., et al. Forced trends and internal variability in climate change projections of extreme European windstorm frequency and severity. Quarterly Journal of the Royal Meteorological Society. (accepted).
3. Koks, E. E. & Haer., T. A high-resolution wind damage model for Europe. Sci Rep 10, 6866 (2020).
4. Gil-García, I. C., et al. Urban Wind: An Alternative for Sustainable Cities. Energies 15, 4759 (2022).
5. Cardona, J., et al. Seeing the Wind: Visual Wind Speed Prediction with a Coupled Convolutional and Recurrent Neural Network. (2019).
6. Higham, J. & Plater, A. `Flowonthego' - flow tracking technology on your smartphone , EGU General Assembly, EGU21-5902, (2021).
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.