Hot climate models and unrealistic assumptions are undermining public trust in climate action
A better understanding of the short-comings of climate models and the unrealistic RCP8.5 scenario is vital for better policy making.
Introduction
Whether you agree or not, many people believe that climate change is a defining issue of our time. Scientific models play a vital role in guiding global understanding and responses. From international negotiations to national policies and local adaptation planning, projections generated by climate models underpin much of the debate.
However, over time, more and more people are talking about their limitations, assumptions, and the interpretations derived from them. In particular, two closely linked concerns have attracted attention:
The tendency of some climate models to predict higher-than-observed warming (the so-called "hot model" problem), and
The widespread use of RCP8.5, an emissions scenario originally intended as a worst-case pathway, as a default or “business-as-usual” projection.
These issues have implications not only for the scientific credibility of climate projections but also for the societal and economic policies based on them.
In this post I examine these concerns from multiple perspectives, including mainstream climate scientists and more sceptical voices. I hope that this offers a rounded discussion for anyone concerned about climate change but open to debate about how we assess and respond to it.
Understanding Climate Models and Scenarios
Climate models are computer simulations that use mathematical equations to describe the behaviour of the Earth's climate system. These models integrate variables such as greenhouse gas concentrations, solar radiation, atmospheric dynamics, ocean circulation, and land use to project how the climate might change in the future. Models also include socio-economic factors, such as population growth, as well as assumptions about future mitigations and land use. Hopefully, you can see immediately that this is a highly complex problem and hence not one amenable to simple interpretations.
The Intergovernmental Panel on Climate Change (IPCC) has used different generations of models (CMIP3, CMIP5, and CMIP6) to produce successive assessment reports. These models do not produce forecasts in the traditional sense. Instead, they generate scenarios based on different assumptions about future emissions and other factors.
I do wonder if you asked a random person in the street, whether they would find it easy to articulate the distinction between forecasts and scenarios? I would argue that most people would likely place more weight on the projections than the models warrant.
Among the most widely used frameworks are the Representative Concentration Pathways s), which represent various greenhouse gas concentration trajectories. RCP8.5 is the highest of these, describing a world with very high emissions, primarily driven by heavy coal use. Others, such as RCP4.5 and RCP2.6, depict more moderate or low-emissions futures.
The "Hot Model" Problem
As part of the latest IPCC process (AR6), the CMIP6 model ensemble introduced many models that produced significantly higher warming projections than earlier iterations. Several of these models estimated equilibrium climate sensitivities (ECS) in the range of 4.5 to 5.7°C, far above historical estimates of around 3°C.
ECS is the amount of global average surface warming expected if the atmospheric concentration of carbon dioxide (CO₂) is doubled and the climate system is allowed to reach a new equilibrium or steady state. Therefore, climate sensitivity is non-linear, i.e. more and more CO₂ is required to achieve the same temperature rise. This is a fundamental feature of the climate system which is poorly understood by the general public.
This raised alarm in the climate science community. Papers such as “Climate simulations: recognize the 'hot model' problem” (Nature, 2022) argued that a subset of CMIP6 models yielded warming projections inconsistent with observational constraints. These models typically had exaggerated cloud feedbacks, particularly reductions in low cloud cover, which amplified warming unrealistically. Again, this is poorly understood by the general public.
To address this, the IPCC chose to apply observational weighting when generating ensemble averages for temperature projections. Nonetheless, the inclusion of these high-sensitivity models continues to influence public discourse and policy analysis, sometimes exaggerating the projected impacts.
Sceptics of the “alarming” climate narrative have challenged the credibility of climate modelling in general. Climate scientist Judith Curry, for example, argues that the persistence of "hot models" undermines confidence in the reliability of future projections. Roy Spencer and John Christy have gone further, suggesting that the climate system is more stable than many models imply, and that observed warming is likely at the lower end of the IPCC's projected ranges.
The RCP8.5 Controversy
Running in parallel with concerns over model sensitivity is the debate over RCP8.5. Originally introduced in 2011, RCP8.5 explores the upper bounds of potential climate impacts under high emissions. It assumes:
a global energy future dominated by coal,
minimal climate policy,
rapid population growth, and
sluggish technological advancement.
Yet, in practice the media and climate advocates too often treat RCP8.5 as the "business-as-usual" or baseline scenario, especially in risk assessments. Analysts from across the spectrum have criticised this. For example, rapid population decline is now likely by the end of the century.
Mainstream researchers like Zeke Hausfather and Glen Peters argue that using RCP8.5 as a default trajectory is misleading, especially in light of current energy trends. They say that as renewables become more competitive and coal use plateaus or declines in many regions, the likelihood of following an RCP8.5-like path diminishes.
I take issue with claims that coal usage is reducing (given both China and India’s increasing consumption) and that renewables are becoming more cost effective (for example see Kathryn Porter’s and David Turver’s substacks). I do see the irony in arguing against this detail when I support their conclusion! Essentially, even without these two underlying assumptions, the RCP8.5 scenario is highly unlikely.
For example, Ross McKitrick, an economist critical of many mainstream climate claims, has pointed out that even with exaggerated emission assumptions, most models still run too hot relative to observations. He calls RCP8.5 a "scare story" used to support alarmist narratives. Roger Pielke Jr., a frequent critic of climate policy orthodoxy, states bluntly that RCP8.5 should no longer be used for baseline policy analysis.
Indeed, energy system analysts note that current global emissions trajectories are more consistent with RCP4.5 or lower, making RCP8.5 an increasingly implausible guide to the future.
Defences of the Current Modelling Framework
While criticisms are valid and often constructive, defenders of the mainstream modelling approach argue that diverse models and scenarios are necessary to capture the full range of uncertainties. They point out that no model is perfect, but the ensemble approach (i.e., looking at multiple models together) provides a robust average that aligns well with past observations when the hottest models are down-weighted. I don’t agree with this argument as I will explain later.
They also argue that scenarios like RCP8.5, while extreme, are useful for stress-testing. By modelling what could happen under worst-case conditions, policymakers can better prepare for high-impact risks, even if the probability is low. This of course assumes that policy makers understand that the probability is low – something I seriously doubt.
Finally, they argue that cloud feedbacks, while still poorly understood, are being continually refined as new satellite data become available. There is also ongoing work to better integrate real-world socioeconomic trends into the Shared Socioeconomic Pathways (SSPs) that now replace the RCPs in more recent IPCC assessments. I would argue that adding in more complexity is just as likely to make the models even more inaccurate.
In short, while the models have limitations, many scientists believe they are improving and remain vital tools for understanding and responding to climate change.
Model Validation - Always Looking Backwards
I question the validity of model validation as currently performed in climate modelling. The standard approach is to produce hindcasts which check how well the latest models follow the historical data.
For example, there is no discussion in the full report for AR6 and earlier reports, about how well previous iterations of models managed to model what happened subsequently. As an experienced modeller, this seems like a gaping hole in terms of model validation. If we don’t have any real measure of how well models project into the future, how much more robust do they become when simply averaged, weighted or otherwise?
At the very least it means that historical policy based on earlier model projections (assuming the models are always improving), has to be viewed with suspicion.
Implications for Policy and Public Communication
The question then arises as to how should we talk about climate projections and use them in policymaking?
First, clarity is essential. Communicators and journalists need to be more precise in describing RCP8.5 as a high-end scenario, not the expected outcome. Conflating it with business-as-usual distorts the narrative.
Second, policymakers should consider using a range of scenarios for planning and weigh them based on likelihood. Risk assessment is important, but not every decision needs to be guided by the worst-case assumptions.
Lastly, greater transparency is needed about model limitations, uncertainties, and areas of disagreement. Demonstrating intellectual honesty is more likely to encourage a healthy and open public debate.
Conclusion
Climate models and emissions scenarios are indispensable tools for understanding global warming. But like all tools, they have limitations. The current generation of climate models, especially some of the CMIP6 "hot models" are known to overstate warming.
Likewise, RCP8.5 no longer represents a plausible baseline, even if it retains value for testing upper-bound risks.
Balancing concern with realism is not climate denialism. It is prudent science and should lead to better policy making. By critically examining how models are used, and striving for transparency in how scenarios are framed, we can improve the credibility and effectiveness of climate communication and response.
For those who are concerned about climate change but wary of exaggerated claims or flawed assumptions, this conversation is not just welcome, it's essential.