Are climate change impacts already included in catastrophe models?

Speak to our team

Interested in finding out more about JBA’s work in climate change? Fill in the form to request a call-back from one of our team.



We’d like to get in touch to find out more about your interest in climate change data. By submitting your details, you consent to the JBA team contacting you for this purpose. We take your privacy seriously. We will securely store the data that you share. We will not share your data with any third party. If you would like to unsubscribe at any time please contact us at hello@jbarisk.com with the subject line Opt-out or call JBA Risk Management Marketing on 01756 799919. All updates will also give you the option to unsubscribe. Read our complete privacy policy here.

As organisations seek to understand the impact of future flood risk, we’re increasingly being asked questions like “How well do cat model outputs reflect the present-day climate?” and “If climate change is already baked into methods, to what extent?”

When considering these questions, we must explore three things:

  1. Whether there is evidence of climate change already influencing model outputs
  2. The trends present in the data and why they are problematic
  3. What modellers can do about these issues.

Is there evidence that climate change is already impacting model results?

To explore this question, JBA undertook some research in collaboration with some of its academic partners, which was published in Kay et al (2018). Together, we used global climate model outputs, a hydrological model, and a Thames basin sub-set of JBA’s UK Flood Model (with climate change removed) to simulate a pre-industrial climate, with the aim of identifying whether changes in climate influenced the number of properties at risk in the south of England during the winter 2013/14 floods.

We did this for thousands of pre-industrial climate realisations and compared them with the actual event. Our results highlighted that greenhouse gas emissions are likely to have increased the chance of flooding in southern England during the 2013/14 winter, with hundreds to thousands more properties at risk during the actual event compared to a pre-industrial climate. This is strong evidence that climate change is already impacting model results, and already influencing property risk. More detail about this research and its methods can be found in a our blog.

Why are trends in historical data an issue?

Many methods used across the re/insurance industry are based on underlying historical observations. There is plenty of research from across the globe showing the presence of trends in river flow levels and rainfall totals within historical timeseries data. For example, here in the UK, a study published by JBA’s Head of Hydrology Duncan Faulkner, Can we still predict the future from the past?, found significant trends in around 25% of the river flow records included in the study (Figure 1). 

Figure 1: Distribution of river flow records with statistically significant trends detected. Mann-Kendall test results for all river flow stations with over 40 years of record (from Faulkner et al., 2019). Triangles show where a significant trend was detected.

Trends in rainfall can be caused by a host of long- and short-term processes, including natural variability (e.g. large-scale ocean-atmosphere patterns such as the NAO and ENSO) and human-induced changes resulting from increases in greenhouse gas emissions. Trends in river flows can result from these changes in rainfall alongside changes in catchments and river systems (e.g. land-use change).

If there are trends present, an estimated 100-year flow level in 1950 might differ from the 100-year flow level in 2020. In fact, some suggest that a model built on data representing a period, for example 1979-2010, represents the climate around the mid-point, i.e. 1995.

Trends in the historical records used when developing methods and models lead us to the question: are these models providing a good representation of the present day, or are they actually reflecting the climate and river hydrology over the past 30-40 year period that the data represents?

What should flood modellers do?

Dealing with trends in historical data is one of many challenges faced by modellers when incorporating climate change into risk analytics. There are a few options modellers can consider to mitigate these issues, including:

  1. Use only the most recent data, which can offer a better representation of the most recent climate. However, the robustness of statistical estimations, especially for extremes, is typically weakened when less data is included. Furthermore, if too little data is used, natural processes that operate over decades might be misrepresented.
  2. Trends could be removed to try and create a more stationary representation of present-day climate. This is challenging as it requires unpicking anthropogenic changes from catchment changes and from natural variability. A range of trend removing techniques has been developed, especially by academia, but these are based on lots of assumptions and often implemented at a local, catchment scale – scaling up to national or even continental scale would be a further challenge. Decisions over whether to leave natural variability in or not would need to be made. This is particularly important if event clustering during a climatic phase should be considered.
  3. Modelled data rather than historical data could be used. Data from climate models is used to make projections of the climate in the future, for example in 2050 or 2080, and 2020 data from climate models could be used to reflect the present day. This approach has the benefit of being able to provide uncertainty bounds as the models can simulate ensembles (multiple representations) of 2020 climates which is vital when considering future risk. However, it is also likely to highlight that the uncertainty associated with climate change itself might be compounded by the uncertainty already present in catastrophe model outputs (especially in the near future).

Incorporating climate science into the re/insurance industry is an evolving part of research and may not hold all the answers yet. Nevertheless, starting to consider the impact of climate change on future flood risk is vital to maintain business profitability under a changing climate.

At JBA, we’re developing a range of climate change analytics to complement our existing UK Climate Change Flood Model, helping organisations to understand potential future flooding and the risk to their business under a changing climate. Get in touch using the form above to speak to one of the team.