By Andrew Pollard, Insurance Specialist at SAS.
A few months ago, the Association of British Insurers (ABI) found that the value of weather damage insurance claims were the highest on record, reaching £573 million in 2023 - a staggering 36 per cent increase from 2022.
The ABI attributed this massive rise to a succession of storms that struck last autumn, including Storm Babet, Ciaran and Debi. Yet, as we reach half way through 2024, this trend shows no signs of stopping, with the UK’s next storm set to mark the most active storm season on record since the practice of naming storms began.
In light of this, actuaries are turning their attention to developing greater certainty in their understanding of climate-related risks, and seeking new ways to improve prediction capabilities to inform policies and pricing.
Technology is paving the way forward here, with AI-powered data analytics transforming the ability to accurately analyse data from various different sources - and IoT-enabled sensors facilitating another one.
Analysing weather data
Data lies at the heart of effective climate risk prediction and management, enabling the development of robust models. But as the climate changes, it’s vital that all data going into models is as up-to-date and reliable as possible - otherwise it risks being a case of ‘garbage in, garbage out’.
One issue here is data silos. There is often a huge amount of historical climate data to analyse, from a variety of government and research institutions, not to mention the latest meteorological data - such as rainfall patterns or river discharge levels, which help inform on flood risk.
AI-powered data analytics is helping to automate and bring together data from disparate sources. It can identify patterns and correlations that may be missed by traditional analysis methods, in addition to flagging any errors or inconsistencies.
Take, for example, a coastal flood risk assessment.
In order to build a comprehensive picture of risk, actuaries may use data sources that include online government flood records, satellite imagery, weather forecasts, topographical maps, and property valuation records. Across the numerous sources this data originates from, measurements including units, coordinates, and timescales may vary, which machine learning (ML) techniques can help standardise.
From this process, patterns will start to emerge, indicating particular combinations likely to lead to flooding. Using this information, the model can help predict the future likelihood a property or area will flood, and estimate what the financial loss could be.
This can translate into real-life preventive action, such as building a sea wall in a high-risk area, or investing in flood-proof building materials. For actuaries, it’ll help determine the appropriate premiums to set when offering insurance, reflecting the varying levels of risk across different coastal regions.
Developments in IoT
As stated, when making weather-related predictions for a company - or a community - it’s important that all data is as up-to-date and accurate as possible. To help make this goal a reality, one town in North Carolina has been using Internet of Things (IoT) sensors to determine when and where flooding might occur.
Prior to implementing its flood-prediction solution, the town’s data resided in multiple disparate systems, meaning they couldn’t get a clear picture and had to approach different places to manually coordinate a response.
Using the SAS flood incident prediction and preparedness solution, powered by Microsoft Azure IoT, they deployed solar-powered sensors and cloud-based predictive analytics to improve real-time situational awareness.
Having all the data pulled together in one place - and the ability to see and analyse everything together - is a major accomplishment for the town. The models are dynamic, continually receiving new information, which helps them understand what’s actually happening on the ground and how to improve it.
With varying amounts of rainfall in different locations, the town can see the timing that houses, businesses, roads and other structures will be affected.
Such a technique would have been unthinkable before the capabilities of IoT, and is an exciting development to those working within the predictions space. After all, one of the best ways to manage climate-related risk is to mitigate it.
Going beyond the traditional
It’s clear that within the climate risk space, there is a need to go beyond traditional methods and adapt, quite literally, to the changing climate.
With this, we’ve seen the rise of parametric insurance, a non-traditional insurance product that offers pre-determined payouts based upon a trigger event, such as an amount of rainfall. The global parametric industry generated USD 11.7 billion in 2021, and this is predicted to rise to USD 29.3 billion by 2031.
The benefits of parametric insurance include quicker payouts, simplified claims processes, and greater transparency. Nevertheless, for the insurance industry, the effectiveness of this depends on accurate and reliable data to set the trigger parameters - proving why IoT data collection and AI risk modelling is invaluable.
We’re now in an era of customer-centric insurance and it is through ML and advanced analytics that actuaries can offer new options which meet customer expectations and help tackle weather-related risk - a win-win.
|
