Governments, regulators and shareholders are increasingly demanding businesses assess and disclose their climate-related risks. To assess the financial risk associated with the physical impacts of climate change, and in particular changes in climate extremes, requires knowledge of climate change across multiple space and timescales. To obtain information on these physical risks, companies are beginning to use outputs from climate models, either directly, or via climate service providers.

While the provision of climate change information, packaged to ease the challenge of understanding the utility of climate projections, seems attractive, the actual limits to the use of climate projections, where these do contain valuable information and where they do not is rarely fully communicated. Companies providing physical climate risk data do not always publicly disclose their methods, or explain publicly, how uncertainty and limits to understanding are managed in the information provided.

Our paper, published in Nature Climate Change (doi: 10.1038/s41558-020-00984-6), documents where climate model projections are useful, where their use can be misleading, and where use is well beyond what climate modellers would deem legitimate. Specifically, we separate the valuable and robust projections of average temperature at sub-continental scales from the projections of extremes, and in particular rainfall extremes at sub-regional scales, which have negligible value in assessing business risk.

The emergence of capabilities in climate analytics presents an interesting challenge to prudential and corporate regulators looking to undertake vulnerability assessments across the financial sector and to place climate-related disclosure on a mandatory basis. Central to the challenge is the fact that climate analytics is an emerging discipline for which a professional body of practice has yet to emerge. So, what guidance can we provide around the use of climate change projections and their use in climate analytics?

A guide to climate analytics

1. Approach climate analytics with a healthy sense of scepticism
The first observation we make is that those who set expectations around climate analytics, such as supervisors, industry bodies and other influential stakeholders, should approach the whole area of climate analytics with a healthy sense of scepticism over the validity of the claims being made regarding numerical quantification. It is a case of asking “how” and of expecting appropriate evidence to be provided. Testing the legitimacy of the approach by engaging with those scientists with core expertise around climate model development would be appropriate. Too often, we see little reference to this type of expertise and no consultation with climate model experts when expectations are being developed for matters like climate-related financial risk disclosure. (The TCFD taskforce, for example, had no climate scientists; nor did the European Union’s High-Level Expert Group on Sustainable Finance.)

2. Using climate models appropriately depends upon the question being asked.
Our second observation relates to why organisations are conducting climate analytics. Are they after a set of defined probabilities of a financial loss from a future event or, alternatively, are they trying to understand how sensitive a particular business model is to a range of possible futures? Climate science can and does provide risk managers and supervisors with useful insights in both of these examples but only through hard work and careful application. There are no “plug and play” answers here and generic products are most likely misleading for any specific application.

3. The black box problem
Given the relative immaturity of climate analytics, it is important that work is undertaken with a high degree of transparency.  We currently do not have the systems of quality control, a body of practice and documented guidance in place that disciplines such as accounting use to ensure comparability, consistency and trust. There would be little point in financial supervisors driving disclosure based upon black-box models or opaque workflows.

4. The paradox of using climate models: difficult, but not impossible.
Our paper could be misinterpreted as advising that no-one should use climate models, outside of the model developers and a few academics with deep knowledge of them. That would be an unfortunate and incorrect reading of our message. In fact, we note that the current research-driven approach to climate modelling is prone to exaggerating the difficulties for business to extract actionable information that is scientifically defensible. We suggest that alleviating this significant barrier is essential. We propose a pathway to doing so by elevating climate modelling to the level of numerical weather prediction, appropriately resourced, organised and delivered in an operational setting supported at one end by climate research and at the other by climate services that extract and provide bespoke information to business and other user communities.

5. Climate models are only useful if vulnerabilities and risks are fully addressed.
Finally, our study also discusses concerns about the overall strategic approach to the use of climate analytics in the first place. This is quite a difficult message and one that is prone to both misunderstanding and distrust. Climate science can only take us so far in the risk journey. It cannot substitute for the development of strong capabilities in the understanding of climate vulnerability by business executives, the development of much better systems to collect and manage risk information and an acceptance that a changing climate introduces a fundamental new risk.  After all, better information about future risks is of limited use if one doesn’t understand or manage the present risks to your business.

Why can’t models just be improved?

We are sometimes asked if new model ensembles with finer-scale resolutions, or more computing power, will soon enable climate models to provide more useful information about business-relevant climate impacts like extreme events. However extreme events are rare by definition, which makes adequate sampling difficult. They are also highly influenced by climate variability (Naturally occurring patterns that influence the weather on seasonal to yearly timeframes)). The amplitude and periodicity of climate variability are not always well-simulated by climate models. Statistical methods such as Generalized Extreme Value (GEV) theory can help in estimating very rare events, however such methods require adequate sample sizes to give robust information on the behaviour of extremes. Various types of large model ensembles may also be used to increase sample sizes, however, GCMs simulate the climate at scales which commonly lack many physical mechanisms that underpin the behaviour of extremes. New initiatives by CMIP6 to increase climate model resolution may help bridge this gap. However, new model issues and biases may arise as models start to include finer-scale processes important for extremes. Moreover, there is evidence that simply increasing model resolution does not improve the simulation of climate extremes (Bador et al 2020). Thus the use of GCMs to assess evolving risks to financing activities is well beyond what climate modellers would consider valid.

Summary

There is a great deal of misuse of climate model projections emerging in business. Climate models are being used for some purposes that are simply inappropriate leading to assessments of the physical risks to business that are of no value. However, there are ways to use climate model data that has value and can help business robustly assess some specific climate-related risks. We recommend several ways forward:

(a) A business needs to ensure that any products that build from climate model projections are developed appropriately. We recommend a business consults experts in climate model development in this assessment.

(b) A business should undertake an internal audit to understand what weather and climate phenomenon affects that business. Once this is understood, a business can ask climate scientists targeted questions relating to business risk and whether model projections can inform these risks robustly.

(c) Generic guides to climate risk, guides focused on averages, or extremes that occur on average every year etc are not likely to inform business risk. Changes in extremes that occur rarely, such as the 1 in 100 year event, are very uncertain and where a business requires this for a specific location, such changes are unlikely to be possible to estimate. That is, it is important for businesses, regulators and shareholders to understand there are limits to what is known about future climate.

(d) It may be possible to improve methods used to estimate physical climate risk, to enable better estimates for business. This will require either bespoke approaches for a business or the creation of an operational environment for climate projections that parallels weather prediction systems. It will take a decade or two to achieve this; businesses that require this level of detail, or regulators intending to require it, should begin communications with governments to establish this capability.

The authors

Tanya Fiedler is a Lecturer in Accounting at the University of Sydney Business School. She has significant expertise in carbon markets and climate change, and in the relevance of these to accounting. Tanya conducts in-depth qualitative research that seeks to understand how, and the associated challenges with, translating engineering and climate models into financial information. Tanya previously worked as a consultant for Energetics, a specialist carbon and climate consultancy.

Andy Pitman is a Professor in climate science at the University of New South Wales and Director of the ARC Centre of Excellence for Climate Extremes. Andy has worked with climate models for 25 years. He was a lead author/review editor on multiple IPCC assessment reports. He has been a member of several NSW and Federal advisory boards relating to climate science. He is a Fellow of the American Meteorological Society and the Australian Meteorological and Oceanographic Society.

Kate Mackenzie is a writer and consultant on the intersection of climate change and finance. She has briefed Australian financial regulators, reporting authorities, corporates and policymakers on these topics, and has held advisory committee positions on several climate science and finance initiatives that span government, research, civil society and the private sector. She is a contributing columnist for Bloomberg Green, and a fellow for the Centre for Policy Development, a non-partisan thinktank.

Nick Wood has 18 years’ experience in assisting businesses manage the risks associated with climate change. Nick is an Associate at Energetics and member of the Stakeholder Advisory Group of the Earth Systems and Climate Change Hub within the Commonwealth National Environmental Science Program). Nick has undertaken analyses of the physical climate impacts to agriculture, and the mapping of information sources and national level capabilities on extreme climate events.

Christian Jakob is the Professor for Climate Modelling at the School of Earth, Atmosphere and Environment of Monash University. Christian has developed weather and climate models for 25 years. He has held several international leadership positions including the Chair of the World Climate Research Programme’s (WCRP) Modelling Advisory Council and as a Lead Author for the IPCC 5th Assessment report. He is a Fellow of the Australian Meteorological and Oceanographic Society.

Sarah Perkins-Kirkpatrick is an ARC Future Fellow at the University of New South Wales. She is a climate scientist specialising in extreme events, with a specific focus on how heatwaves have changed in the historical record, and how they will change in the future.