Climate scientist and mathematician Dr Emily Shuckburgh talks to Chris Seekings and Sharad Bajla about modelling the future impacts of climate change, and how actuaries can help
The overwhelming majority of scientists agree that man-made climate change is real and threatens to radically change both the environment and society. Rising sea levels, frequent extreme weather and vanishing ice sheets are just some of the impacts predicted under a business-as-usual scenario.
Dr Emily Shuckburgh, mathematician, climate scientist and director of Cambridge Zero, explains how these forecasts are made, the challenges behind designing accurate models, and the role actuaries can play in helping predict future outcomes.
First, the facts
Cambridge Zero brings together all the engineers, scientists, lawyers and social scientists responding to climate change at the University of Cambridge, with Dr Shuckburgh responsible for coordinating this effort. Although many unknown variables go into forecasting the scale and urgency of the issue, she first outlines various “observational facts”. “We know from the ice core records that over the last almost million years, carbon dioxide levels have ranged between 180 and 280 parts per million in the atmosphere,” she explains. “It’s observational fact that carbon dioxide levels are now well above 410 parts per million, grossly exceeding historic levels.”
Scientists say temperatures are rising thanks to the greenhouse effect, with gases trapping heat in the atmosphere and preventing it from escaping into space. “It’s also observational fact that the average temperature is now about 1.1°C warmer than it was 150 years ago, before the Industrial Revolution was really starting to ramp up. We know that we’ve changed the atmosphere in terms of its chemical composition and that has had an impact in terms of temperature.”
The volume of water in the ocean expands when heated, and ice that was once on land finds its way to the ocean due to melting; in this way, sea levels have risen by around 25cm since the 1900s. Perhaps the most impactful change scientists have observed, though, is the frequency and intensity of extreme weather events such as heatwaves, flooding and wildfires. “Whenever there is an extreme weather event, the scientific community systematically asks: ‘Has climate change increased the risk of these events occurring?’” Dr Shuckburgh says. “In the last couple of years, we’ve been able to say that that risk has increased, and in many instances, increased manifold – that’s been the current impact of climate change.”
Sea levels have risen by around 25cm since the 1900s
Making models
She explains that, to forecast future impacts, scientists first have to appreciate the physics of the climate system. “Both the atmosphere and the oceans are fluids sitting on a rotating sphere that is subject to gravity,” she says. “Those are the key inputs if you want to build a model of how the atmosphere or oceans evolve. Just writing down the physical equations describing that is at the heart of every weather forecast or climate model.”
There are, however, other key inputs when predicting the climate, such as the equation for how energy from the sun moves through the atmosphere. “We need to understand how energy from the sun, which is shortwave radiation, moves through the atmosphere compared to the longwave radiation that is reflected back from earth,” she explains. “The difference between that is the greenhouse effect.”
There are also limitations to how much detailed data scientists can put into the model. Processes behind the evolution of sea ice and formation of clouds cannot all be directly included in computer models. “We have to use statistical representations of them instead and look at the patterns of change,” Dr Shuckburgh says. “It’s a bit like the resolution of a digital photograph. The higher the resolution that we have, the more information we have, the greater fidelity of the model, but there’s a limit because we have limited computer power.”
Looking back
Actuaries use a process of backtesting when modelling life expectancy and other future outcomes, comparing previous predictions with what has actually happened. Climate scientists use a similar process. “We will develop our models and then run them over the last century and compare them to actual temperature changes. We will then also test models completely out-of-sample, and look to see they have been able to replicate very different climates.”
Climate scientists test models in so many ways, and there are so many aspects to the climate system, that improving these models is “often very, very complicated”. “It might be that there’s a systematic difference in the temperature profile in the South Atlantic in the models compared to the real world,” Dr Shuckburgh explains. “That might be because clouds have been misrepresented in the region, which is having a knock-on impact on oceans. Trying to trace back how you can improve one aspect of the climate model involves looking at how the model is represented through the physics.”
Data from satellites is now available to scientists on a daily, or even hourly, basis, and modelling centres around the world each have marginally different ways of representing the physics. “They’re all slightly different and come up with slightly different projections that then give a further range in terms of our predictions,” she says. “Of course, the other key thing that is uncertain is what our future emissions of greenhouse gases will be.”
The average temperature is now about 1.1°C warmer than it was 150 years ago. The world is on course to exceed 3°C of warming by the end of the century
Known unknowns
Scientists don’t have one single prediction for the climate. There are a range of different scenarios, and it is difficult to incorporate what Dr Shuckburgh describes as the “known unknowns”. These could include the collapse of vast ice sheets covering Greenland and West Antarctica, the rapid dieback of the Amazon rainforest, or the melting Arctic releasing vast quantities of methane, a powerful greenhouse gas, into the atmosphere. “There are a set of high-impact but low-probability aspects of the climate system that we know might occur, but it’s difficult to incorporate those formulae within predictions – West Antarctica has three meters of sea-level rise equivalent within it.” Dr Shuckburgh explains. “Then, of course, there’s a whole set of unknown unknowns that we have no possibility of being able to include. As our scientific understanding evolves, then more and more things get moved from the unknown unknowns into the known unknowns and so forth, helping quantitative understanding.”
Climate scientists look for trends, rather than isolated changes, and another potential known unknown involves the international pledges made by countries to cut emissions. Even when these commitments are incorporated into models, the world appears to be on course to exceed 3°C of warming by the end of the century, warns Dr Shuckburgh. “That’s the state of play, and then we need to understand what the implications of those temperature rises are.”
This area could be of particular interest to insurers who might be looking to devise flood or fire policies. “We’re seeing an increase in flood risk because a warmer atmosphere is able hold more water vapour, which means the air is wetter and more prone to heavy rainfall events,” she says. “Damage caused by tropical cyclones tends to be through storm surges, and with rising sea levels, storm surges are able to penetrate much further inland. We’re very much seeing an increase in the frequency of heat waves in some few parts of the world, too, and that could be a contributing factor to wildfires.”
Machine learning
One way that scientists have been trying to quantify known unknowns is through artificial intelligence (AI) and using machine learning algorithms with strong mathematical underpinnings, including to automatically interpret satellite data. This can help not only with climate forecasts, but also with predictions about the world generally. “It might be being able to count the number of a particular species of whale from space and to see how those numbers are evolving in the context of climate change, or the evolution of unplanned settlements around growing cities in developing countries.”
Air quality sensors and wearable technology can also be used to inform AI about indoor and outdoor air pollution as people move around in their everyday lives. “We are integrating climate-related datasets with very different epidemiological datasets to understand the relationship between different weather conditions and health outcomes,” Dr Shuckburgh continues.
“We can then use those two together to try and predict how climate change might alter our health outcomes.” We speak to Dr Shuckburgh as many countries go into lockdown amid the COVID-19 crisis. “Trying to understand the ways in which climate change might impact how different diseases transmit between humans, animals and plants and so forth is a key area of research that many people are involved in.”
She says that coronavirus could provide some relief for the environment in the short term as restrictions on air travel and falling oil demand bring down global emissions. “It is a very direct and immediate impact just in terms of this year’s CO2 emissions – I suspect this year they will be low because there’s lockdown,” she continues. “How that plays out in the longer-term is very uncertain. If this crisis can be used as an opportunity for restarting the global economy on a cleaner trajectory, then that might be at least one fruitful outcome from what is currently very much looking like a hugely concerning global crisis.”
Actuarial opportunities
As director of Cambridge Zero, Dr Shuckburgh understands the importance of drawing on expertise from various different disciplines, and always wanted to use her mathematical background for causes outside academia. “Mathematics is very beautiful in its own right, but I was very keen to use maths for useful real-world purposes,” she explains. “I started being involved in climate-related research right at the start of it becoming an international political topic, and it’s been a convergence of different interests.”
The subject is of increasing interest to actuaries, too, as it becomes clearer how climate risks threaten to impact health, weather and the global economy. Dr Shuckburgh believes there are “huge opportunities” for actuaries who wish to apply their skills in this area. “Something which is very relevant to people with a mathematical background is understanding how machine learning and data science methods can be applied,” she continues. “There are a lot of exciting opportunities there. My work is involved with quantifying risk in different dimensions, and I think a huge number of opportunities are at the interface of climate and actuarial studies.”