Modeling extreme heat waves

Welcome to Science Sessions, the podcast of the proceedings of the National Academy of Sciences, where we connect to with Academy members, researchers, and policymakers. Join us as we explore the stories behind the science. I'm Paul Gabrielson. Climate models are very useful for reproducing climate processes, especially processes we understand well. But there are some processes that models struggle to capture, like unusually rapid increases in the frequency and intensity of heat waves. In a recent PNAS study, Kai Cornhooper of the International Institute for Applied Systems Analysis in Vienna, Austria and colleagues sought to understand how well climate models represent extreme heat wave events. They found hotspots around the world where extreme heat waves are increasing in ways climate models currently fail to reproduce. Kai, let's start with some context. Why is it important to be able to accurately model heat waves? Heat waves can cause a lot of harm. They're extremely dangerous to the human body. They cause a lot of excess mortality. They're also risky for infrastructure and energy systems and agriculture. With ongoing global warming, we already have seen an increase in the frequency and the magnitude of heat waves around the world. We are going to expect more frequent and more intense heat waves in the future of greenhouse gases increase so it is intensely important to be able to model them accurately, to be prepared for what is to come under different climate scenarios. How did you identify hotspots of faster warming? We were interested in investigating this tail behavior of the heat wave distribution. So if you imagine a quasi-goshian distribution of temperatures, we were interested to see how the most intense values of that distribution are going to change with respect to the more moderate ones. So we coined this metric tail widening where we would investigate the changes in the most extreme heat events and compare them with the more moderate ones. That difference we mapped around the world to highlight and identify regions where these tails are accelerating faster than what would be expected from just a linear change in temperatures associated with global warming. It allows to highlight areas where more complex non-linear processes might come into play to drive heat wave trends and these could be changes in the atmosphere dynamics or specific processes associated with dry soils and warming. These types of feedbacks are known to accelerate heat waves and really drive the intensities above what has been observed earlier. Where are these hotspots? We identified several areas around the world in particular Western Europe popped up as a heat wave hotspot but we found other areas almost at every continent globally. So south and south America is a hotspot as well. Australia and the southeast has large regions where the hottest temperatures increase faster than the more moderate ones and also northern Canada and Greenland. When we do understand the physical processes that lead to this access trends and extreme heat we can do some statements about what might happen in the future as well. For Europe several studies have suggested that changes in the mid latitude large scale circulation jet stream might have driven these temperature extremes in the summer and several studies have looked into how this is going to change in the future. So in the case of Europe we can say that if the greenhouse gas emissions continue to rise we expect more extreme heat waves and possibly also more extreme than what climate models have predicted. How well do climate models represent these rapidly warming areas? Climate models are remarkable tools and they have been excellent in reproducing our observations on the global scale and also more moderate extremes and regional temperature responses are really well reproduced by climate models. But of course the more regional you want to be and the more extreme your hazards become the more challenging it is for climate models to reproduce these types of phenomenon when comparing the observed trends on a global scale with what climate models have predicted will happen. For these most extreme trends climate models fail to reproduce the observations and in some cases the set of climate model simulations that we looked at were not able to capture the observations at all and that of course is an issue if you want to make accurate predictions about future climate risks. The more extreme the events are the further you go into the tale of the distribution the harder is for a model to reproduce this. These extraordinary extremes they are the outcome of various processes acting together on different temporal and spatial scales including atmospheleonetics, humidity, local geomorphology and so on. So it becomes fairly complex to accurately model these events. Our task as scientists now is to raise awareness of this or the finance sector the insurance sector just to name a few are extremely interested in these estimates and they make decisions that affect a lot of people's lives and if these estimates do have a systematic bias towards being too conservative or too low then this is going to cause a lot of issues in the future. So I think that is the first purpose of the study. The second one is to motivate scientists to identify the physical processes that drive these trends and then to find a way to incorporate these processes into climate models to make them better. What can climate models do to improve representation of heat waves? First we have to identify the processes that drive these heat waves on different temporal and spatial scales and an obvious way is to increase the spatial resolution of the climate models. Another way would be specific machine learning approaches to include processes that cannot be explicitly integrated as equations but as parameters and these parameters can be fitted or adjusted with new methods from within AI. One issue of course with extremes is that they are so rare. So if you have only a few model runs you might be unlucky and they just do not include any extreme for stochastic reasons. So running a model that you trust a lot of times to really map the distribution well is also crucial in that respect. What are the caveats or limitations of the study? First of all we look into observational data. So observed trends can be due to anthropogenic climate change and rising greenhouse gas concentrations in the atmosphere but there might be a small component that is also associated with natural variability that can drive temperature trends on a regional scale and that needs to be investigated to what degree these factors count into these trends here. Another caveat is that we rely on reanalysis data which is based on various observational products which are then assimilated into one globally gridded data set but themselves use the help of climate models. Reanalysis data sets themselves have a bit of a bias and observations usually show more accurate. So the reason why we relied on reanalysis data is because it's globally gridded and we can cover really every spot in the world but to be really sure about the degree to which these tails are widening one would need to look into observational data sets on a global scale. What are the major takeaways from your study? These types of predictions that we really rely on they need to be considered conservative estimates. They are usually pointing us towards a right sign but they're often not of the right magnitude in terms of trends especially if you're a city planner or if you're a farmer or if you're interested in how your climate future will look like and you rely on these types of models and they are used in a lot of risk assessments. You should assume these types of estimates rather conservative values. I think that's a key takeaway. We're entering in domains where it becomes really difficult to accurately model risk and and project risk. So we're entering in these new types of climate regime. that we are not prepared for right now. If you are a responsible policymaker and statements about future risks and what to expect in the future are increasingly difficult to make and increasingly associated with more uncertainties. (upbeat music)