By Dr. Joshua Woodbury, Senior Climate Risk Scientist
A recent article in Nature, titled “Business risk and the emergence of climate analytics”, has created a lot of buzz around the use of climate models in the financial sector and is worth reading. It points out some of the potential pitfalls of using general circulation models (GCMs) in business decision making, particularly around projection scales (global vs regional vs asset level), time frames (the next few decades vs the end of the 21st century) and extreme events. The authors have many good points, and the article provides us with a useful ‘pump the brakes’ moment, to think about how we are using these models (and even if we should be at all).
“All models are wrong, but some are useful” is a popular aphorism generally attributed to the statistician George Box, but most modelers will know this innately, whether they’ve heard the saying or not. This is because models are necessarily idealized versions of the processes they try to represent and always fall short of the complexities of reality. The question should never be ‘is the model correct?’, but rather ‘is the model useful?’. This is particularly the case when using GCMs for business decisions, as they are idealized versions of the physical world, not the business world. So, are these GCMs useful for understanding the potential physical climate risks to the financial services sector? We say yes (and the authors of the article seem to think so too). While GCMs were not directly designed to answer many of the questions that the financial sector is asking around physical climate change risk, they were designed to understand the physical changes due to climate change, which we think is a good starting point. However, the GCM outputs need to be well understood and more often than not combined with additional modeling to gain insight that is useful to the financial services sector.
The major theme throughout the article is the potential misuse of the GCMs for the purpose of making business decisions around climate change. While the limitations and uncertainties of these models are numerous, we think the key (as it appears the authors do too) to their usefulness is having a deep understanding of the limitations and uncertainties. For example, the article rightly points out that the models are currently not sufficient for understanding how extreme events will change into the future (a relevant question for the financial sector). However, this does not mean that GCMs are useless here. While GCMs do not explicitly resolve the dynamics of tropical cyclones, key variables such as sea surface temperatures, potential energy and wind shear can be investigated under different scenarios. Of course, these variables do not explicitly tell us whether a tropical cyclone will occur, they do give us an indication as to how the frequency and intensity of tropical cyclones may change. This is the approach we take when estimating the impact of climate change on tropical cyclones. Our tropical cyclone model is primarily based on extreme event modeling methods (comparable to the natural catastrophe modeling approach used widely in the insurance industry) using today’s climate and historical events. Future tropical cyclone scenarios are based on how key governing variables may develop in various GCMs, following Knutson et al (2020). The GCMs are not directly generating our tropical cyclone impact estimates, but they are key to the process. There may be a time in the future where these models can resolve extreme event dynamics and can efficiently be run many thousands or hundreds of thousands of times to resolve extreme statistics, but we aren’t there yet. Until then though, there are ways to make the GCM outputs useful here.
While our internal modeling requires expertise and a deep understanding of the dynamics and processes, the most important part for us is transparency in communicating the uncertainties and limitations to our users. We believe that the most effective way to have truly informed, longstanding users is to be transparent about the limitations, uncertainties and assumptions involved in this new and complex area of climate risk modeling. It benefits no one to create black box models for uninformed users, as this will likely only lead to frustration and a superficial treatment of climate risk. Neither of which will help financial institutions understand, quantify, and respond to the climate risks that could dramatically affect their business models. At this early stage in climate risk modeling, the models should be used as guides, not answers, to help understand how the future might unfold under different scenarios. This is only possible with transparent modeling and truly informed users.
The article says that “risk disclosure and decision-making across many levels of economic activity has leap-frogged the current capabilities of climate science and climate models by at least a decade.” While it’s difficult to say exactly how far behind these models are, it is true that the GCMs as they are do not directly address current risk disclosure and decision-making processes. However, this doesn’t mean that they are entirely useless. We just need to be clear about the limitations and uncertainties of the GCMs and how they are used, as the authors continually stress. While we generally agree with the points made in the article, we must not let perfect be the enemy of good as financial institutions begin to act on a changing world. Climate risk modeling for the financial sector will continue to evolve and we at Planetrics are committed to supporting the industry on this journey by putting forward the best available scientific insights.
About the author
Josh has a background in physical risk modeling, insurance and climate change. At Planetrics he combines these skills to develop and deliver cutting edge climate change analytics and tools for our clients. Email Josh at Joshua.Woodbury@planetrics.com.
Please fill in the form and the team will get back to you by return. Alternatively, use any contact method below.