Member Spotlight: KatRisk
Dag, what is your background and why did you found KatRisk?
I originally studied physics and conducted my PhD in climate science – specifically on how to combine hydrology with climate and weather prediction models. I worked at Princeton University for a few years and then for various US government weather and environment agencies. I then realised that neither academia nor government was my future; I wanted something more entrepreneurial.
I worked for RMS where I built continental-scale models for Germany and the UK, as well as simulation methodologies. After almost eight years, I left the company to build something of my own with two former colleagues and friends. We created KatRisk due to the opportunities we saw in the market – at the time, there was a lack of high-quality global flood models.
What does KatRisk offer to the insurance industry?
KatRisk provides probabilistic models, hazard maps and portfolio analysis software. When we began in 2012 we focused on inland flood maps, as that was what I had heard potential clients asking for. By 2016 we had created a probabilistic tropical cyclone wind and storm surge model, which we combined with our inland flood model. We used this to provide clients with footprints after Hurricane Harvey made landfall in 2017. We won FEMA (the US Federal Emergency Management Agency) as a client after this.
We now offer catastrophe models in the US, Canada, Europe and Australia, with additional regions currently under development. We also provide hazard data and location loss analytics for hail, tornadoes and wildfire in the US, with a resolution of up to 10 metres. Data is developed in real-time for significant natural catastrophes to assist insurers with post-event assessment, like our global flood risk forecast system: floodriskforecast.com.
A very important part of our work is the software and data infrastructure we have built. Our high-resolution models and the financial model have been built to offer very fast runtimes. We have achieved this by optimising all components in the models, such as data compression, parallelisation and fast simulations with the best possible loss convergence. We have also made sure that our customers can see all the data transparently and that we have an auditable financial model. We made sure to always have coherent risk measures (like additivity) when running our models – for example, our year loss tables are additive when a portfolio is split into two smaller portfolios.
How does KatRisk incorporate climate data into its models?
We realised very early on that we needed to incorporate a climate change component into our models. This includes sea level rise and globally increasing temperatures, among other variables. This has always been a feature of the models we build, allowing us to see how storm surge risk, for example, changes over different warming scenarios up to 2100.
The other component we added to our models was climate variability. We create global correlations between all models firstly by modelling 50 thousand years of global sea surface temperature. This allows us to explore the impact of the El Niño–Southern Oscillation (ENSO) or changes in the Atlantic sea surface temperature on the likelihood of extreme events. We can then investigate features like seasonal predictions and event clustering within our models.
What inputs go into creating KatRisk’s models?
When we create a catastrophe model, the majority of data we use is open source. The data usually comes from government agencies globally including NOAA, NASA, ECMWF and weather services in Japan and Australia. To translate this data into a model that insurers can use to make decisions, a lot of insurance industry knowledge is needed – not just expertise within climate science.
Another important part of creating catastrophe models is knowing what not to do. It is important for us to reflect the most important aspects of catastrophe models, such as frequency, severity and correlation. There is a risk that one can build models that could be made arbitrarily complex – something we are trying to resist.
How do clients access KatRisk’s models and data?
There are four different ways to deploy our models and data to clients.
- The first choice is for a customer to licence all of our data (which is about two terabytes). We send it to them on a hard drive. They can then install it on any server, workstation or laptop.
- The second option is for our models and data to be installed in the cloud, using service providers such as AWS or Azure.
- The third choice is for Katrisk to host the data on the cloud for a client. In this case, we would rent a server and deploy it for them.
- The final way of accessing our data is through third-party resellers. This is usually just for our hazard data and can be retrieved via an API that we provide.
If a potential client would like to test our data and models before making a decision to licence us, we can also run a bespoke consulting project for them.
How can KatRisk support clients with ESG reporting?
There are a lot of data providers out there that only provide maps with a colour scale of how the climate is going to change. For example, if an area is going to be affected a lot more by heatwaves, it would be marked as red for this peril. In many cases, this offering is just climate data downscaled, with not too much thought to the practical impact of these climate issues or how extreme events might change. In my opinion, companies often use this data solely to comply with regulatory reporting requirements.
Insurance companies have gained a lot more expertise in climate risk – they have analytical teams that can delve into climate science to understand more about it. At KatRisk we provide data on how climate and extreme events will change in the future. We are currently exploring how we can help financial institutions with ESG reporting in the most effective and principled way.
What clients does KatRisk work with?
Most of our clients are based in the US and Europe. These clients often have interest in our maps and data outside of these regions too, such as our Australia or China flood maps. Four out of five of the top global (re)insurers are clients of ours, as well as the top five largest brokers. In total we have more than 60 clients.
We have a range of other traditional and start-up insurers as customers. Outside of this, FEMA is an important client. KatRisk has been used as the calculation agent for the six catastrophe bonds they have supported, where over $2 billion of flood risk has been transferred to capital markets.
What companies would KatRisk like to connect with?
KatRisk is looking to work with insurance companies that need a sophisticated look at global flood risk, especially when correlated to tropical cyclone wind and storm surge risk. We also offer a different view of risk on perils that are covered by the large modellers Verisk and RMS, such as US severe convective storm models.
We are also looking to work with academia or those who are trying to create new risk transfer solutions. This can include, for example, parametric catastrophe bonds.
What should readers do if they want to learn more?
Our website is a good place to go. You can learn more about our software, hazard data and financial models from there. To hear our thoughts on recent hurricanes, you can watch this video.
I would also love to hear from anyone interested in working with us directly – they can email me at [email protected].
