Weather Data and its evolution with the Insurance Industry
In 2019, the Association of British Insurers reported that UK Property Commercial and Domestic Insurance Weather Claims comprised 11% of the value of gross claims incurred, equal to an annual total of £642.5m and a daily total of £1.76m. The majority of weather claims are linked to severe weather events such heavy rainfall or high winds, but attritional damage from the consequence of sustained weather conditions is a consistent issue.
In August 2021, the reinsurance group Swiss Re commented that the first half of the year was characterized by ‘ever higher natural catastrophe losses’ due to extreme weather caused by climate change and the consequences of ever-increasing urban development. The rising costs to the insurance industry and the underperformance of the property and casualty underwriting result brings increasing efforts to examine ways to better harness the growing availability and scale of weather data and computational resources.
Developments in technology, data and processes have evolved the nature of the conduct and output of weather forecasts and observations. Concurrently, the experience of Covid-19 has increased the expectations of the insured customers for mature, personalised digital engagement. The insurance value chain is steadily acknowledging the growing opportunities to harness weather and climate data for mutual benefit, but several challenges remain to be overcome.
Mr James Burn is a Meteorologist with the IBM Weather Company and has over 20 years of both commercial and academic experience. He specializes in helping organisations flourish with weather data, particularly in a digital environment. Mr Burn is especially focused on imparting knowledge on how weather modelling works, the fundamental importance of data assimilation and quality control, the methodology of multi-model forecasting and achieving the most possible from computing resource. Balkerne speaks to James Burn on these topics as they discuss the growing importance of weather forecast and observation data to the property insurance industry.
Balkerne: Could you explain the difference between weather and climate?
James Burn: They are closely linked, but time is the key difference. Weather is a short-term description of the atmosphere – variables like sun, rain, temperature, wind etc. Weather timescales cover recent observations and forecasts hours to months ahead. Climate covers the statistical analysis of weather over longer time periods. This is often thought of as “average weather”, but climate analysis can involve sophisticated statistical methods.
Balkerne: Over the past five years, what do you think have been the most significant changes and developments to weather forecasting and the conduct of weather observations?
James Burn: Great question. Meteorology continually changes and improves with new technology – often from areas outside the core science. Here are 3 major changes:
1. Probabilistic weather forecasting (my favourite topic) is increasingly used by organisations making weather influenced decisions. I wonder if raising the profile of data analysis in the Covid-19 era is helping the public better appreciate statistical based recommendations?
2. A technique called GNSS radio occultation uses measurements between low earth orbiting satellites to derive profiles of upper atmosphere data. Upper air observations have previously relied on a few hundred weather balloon releases; incorporating radio occultation derived weather data into models gives significant increases in accuracy.
3. Improvements in computing resource, essentially bigger computers and the intelligent use of computing resource is allowing weather models to run at higher resolution.
Balkerne: From your experience what is the relationship between insurance claims and types of weather events?
James Burn: When I first worked as a meteorologist, we collected various weather data together to prepare “hindcasts” and help validate insurance claims – this was a very manual process. Nowadays, I think it’s easier for insurance companies to access historical weather data. However, the meteorological science of choosing and combining the data remains. I do feel insurance companies are adopting to the improved data available. There are now seasonal weather forecasts feeding into claims models, and new climate risk models – these show a level of accuracy enough to get the attention of some insurers!
Balkerne: What is the role weather data can play in claims efficiency, both in validation and fraud mitigation but also with the concept of parametric insurance?
James Burn: It’s very important. For claims validation, we have insurers starting to use data from personal weather stations, alongside the traditional trusted government sources. With some good quality control algorithms, it is possible to achieve value from these unofficial sources.
For parametric insurance, typically it’s been about providing historical datasets to build contracts and trigger automatic payout. As data accuracy increases, there is increasing interest in weather forecast data (probabilistic and seasonal) to build into parametric contracts. This would help trigger payments before an event. So you’d get an automatic payout days before the storm hits, or potentially months before a drought ruins your crops.
Balkerne: From your experience of organisations conducting mature digital engagement and issuing tailored alerts to customers prior to a forecast extreme weather event, what have been the outcomes?
James Burn: We work with a UK insurer who use weather triggers to deliver timely warnings to policy holders. This has been well proven to increase customer retention – insured customers appreciate the warning and the advice. Actual claims reduction based on weather alerts is an evolving and growing opportunity but it is too early to see a definitive statistical improvement. There needs to be more ownership by companies and the public of the problem, together with pro-active behaviour.
Balkerne: What opportunities do you think exist for insurance stakeholders, because of the availability of a greater geo-temporal frequency and accuracy of weather forecasts and weather observations?
James Burn: There are many opportunities. We’ve covered some of them above – weather alerts to increase customer retention, parametric insurance payouts before adverse weather strikes, improving claim validation and fraud detection. Many recent improvements in meteorological science are based around data and statistics. The insurance industry has statistics and data at its core, and I feel that the weather and insurance relationship will generate many more opportunities in the coming years.
Balkerne: Final question, is it going to rain tomorrow?
James Burn: Yes, it will certainly rain - somewhere! It’s good to trot that one out from my forecasting days…