Picture the scene: it is 2050 and you find yourself Chief Scientist at the bridge of the Planet Earth Climate Monitoring Centre. You look around your Virtual Reality Earth Information System (VIREIS) and observe with some satisfaction that the rate at which carbon dioxide concentrations in the atmosphere are increasing has slowed considerably over the past twenty years, as has the rate of increase in global average temperatures – especially over the last decade.  

You zoom in to take a closer look at a few regions. There is a cold spot over the North Atlantic, which your system confirms has been caused by a 30% slowdown in the Atlantic Meridional Overturning Circulation since 2020.  The ocean around Antarctica remains worryingly warm; on the other hand, the West Antarctic Ice Sheet is melting more slowly and appears to be more stable than some had feared.   

From the Turbulent 2020s to a Global Turnaround

In the past year – typical of almost every year now – there have been record-breaking extreme events on every continent, some shattering previous records. Heatwaves have become a particular menace, taking lives and destroying livelihoods. Unprecedented droughts, wildfires, and floods have also proved extremely damaging, although innovative adaptation measures have greatly reduced the damaging impacts on people and Nature. Indeed, the progress in effective adaptation, alongside the successful mitigation of climate change enabled by the remarkably rapid energy transition, has been a major success of the last two decades. This was a big turn-around, following the tricky 2020s. Ultimately the collective demand from global public opinion, aided by the compelling economics of renewable energy, brought about the changes required.

You turn your attention to the future. The system shows how temperatures and many other aspects of climate are expected to stabilise over the coming decades. Global sea level rise will continue for much longer, of course, but at a rate much slower than had been feared, enabling many more options for coastal communities. All things considered, the outlook is considerably brighter than you had expected 25 years ago when you began your career in climate science.

How Climate Science Keeps Us Informed and Prepared

This isn’t, of course, quite how things will turn out. But how much of it might prove accurate? In particular, how will climate science, and the role of climate scientists, have changed by 2050?

Let us start with considering how the world will have changed. Unless there are major unexpected events (e.g. a series of major volcanic eruptions, very significant climate intervention) the climate of 2050 will be warmer than that of today, possibly by up to 0.5oC or even more, with all the associated consequences that this entails¹. There will, therefore, be strong ongoing demand for climate science, data and intelligence from governments, businesses and people:

• To quantify the state of the climate/Earth system and the rate of ongoing changes at global, regional and local scales.

• To understand and explain (“attribute”) the causes (typically multiple) of specific observed changes, events and impacts, and their implications for the future.

• To anticipate future changes in the years and decades ahead, to inform risk assessments and investment and planning decisions.

• To provide early warning for extreme and high impact events, so that lives and livelihoods can be protected, and for abrupt or potentially irreversible changes in the climate system.

• To design and evaluate options for reducing future climate change through mitigation measures, and for reducing the negative impacts of climate change through adaptation and building resilience.

These demands will exist in some form in every country and community in the world, albeit always with specific local needs and priorities1,2.

AI and the Future of Climate Modelling 

A next question is how will the scientific capabilities to monitor, model and predict climate have developed by 2050? The likely most important area of technology innovation concerns the role of Artificial Intelligence and Machine Learning (AI/ML). To begin with monitoring, we can be confident there will be an enduring need for both remotely sensed (notably from space) and in situ observations. The growth in observations from autonomous platforms is likely to accelerate, as is the role of diverse networks of distributed sensors. AI/ML developments will provide new ways to extract the maximum and most decision relevant information from a likely unprecedented diversity of measurements.  We can also hope they might provide a near complete picture of the current state of the climate system – the basis of our imagined VIREIS – including timely evidence on the impact of climate actions, both for mitigation and adaptation.

In the area of modelling, AI/ML developments will also be very influential but (along with many other climate scientists) I anticipate an indefinite requirement for physics-based climate and Earth System models as well. The reason is – because future climate cannot be observed – physics-based models and associated process understanding are fundamental to confidence and trust in longer term climate predictions and projections.  (And the importance of trust should not be underestimated.) Putting this another way, we will never have sufficient observation-based data to adequately train a purely data-driven climate or Earth system model to produce reliable long-term projections; such models perform poorly when pushed far outside the applicability of their training data.   But this isn’t to say that AI/ML approaches cannot add a great deal of value to climate modelling.  For example, they are already showing potential value for problems such as: novel parameterisation of unresolved processes; extended range initialised predictions³; model calibration and tuning; downscaling⁴; and ensemble generation.  For many applications, hybrid physics-ML approaches are likely to offer the best solution.

AI/ML approaches will also have a central role to play in enabling the maximum possible benefit to be derived from the computational resources of the day.  The allocation tension between resolution, complexity, duration, and ensembles will not be solved, but it should be helped.  We can anticipate further advances in the resolution of physics-based models, but many processes will remain resolved only partially or not at all, and – unless there is a major and unexpected technological advance – these simulations will remain expensive so will have to be useful carefully and selectively.  The same is true for simulations of the more complete Earth System including the carbon cycle and other processes important to the longer-term evolution of climate (e.g. ice sheets). AI/ML advances should help us target these expensive simulations so they add most value.  For some applications AI/ML based emulators will likely suffice, although these will have to be trained at least in part on traditional physics-based model simulations.  There are interesting ideas already in development about the potential for constructive co-evolution between simulation and emulation capabilities⁵.

Another feature of VIREIS is that it will exploit AI/ML methods to combine observational and model data to provide a seamless integrated capability for monitoring, attribution and prediction of climate on all scales from global to local.  This will realise the vision of the WCRP Lighthouse Activity on Explaining and Predicting Earth System Change⁶. It will also provide early warnings both of individual extreme weather events (and their impacts), and of the potential for rapid, unexpected, or irreversible changes in the climate system, including possible tipping points⁷.  

The next area of climate science application is the design and evaluation of options for mitigating (i.e. reducing) future climate change, for example by cutting greenhouse gas emissions through accelerating the energy transition or through changes in land use.   To meet this challenge, climate scientists will be working closely with experts and data from many other fields including engineering, economics and behavioural science (building on current initiatives such as https://ukncsp.org/).  Data and models may be integrated into new forms of digital twin that enable a range of future options to be explored and evaluated considering a wide range of metrics including co-benefits and trade-offs.  Also relevant here are potential developments in climate intervention8, both Carbon Dioxide Removal (CDR) and the more controversial Solar Radiation Modification (SRM) approaches.  How these approaches develop, and whether there is pressure for them to be deployed, is highly uncertain.  

By 2050 (and hopefully much sooner!) adaptation to climate change, or more broadly “designed climate resilience” will be a mainstream feature of all major investment and planning decisions.   Here again climate scientists will be working with experts and data from many other disciplines to co-design and co-evaluate resilient solutions to challenges as diverse as: energy infrastructure, transport networks, health services, water supply, agriculture and food production, and global supply chains. AI innovations will no doubt enable development of much more powerful and user-friendly decision support tools.

Another dimension of the 2050 landscape we might briefly consider is the future of international institutions such as UNFCCC and IPCC.  Notwithstanding current challenges, it seems clear that the need for something like UNFCCC is not going away.   It will no doubt evolve as it has done before (e.g. bringing in new mechanisms such as the Global Stocktake) and perhaps change in more radical ways. One area where I would like to see more attention is the need for a comprehensive and regular global assessment of the climate risks that are avoidable through effective mitigation action.  It seems absurd that the world does not have such a resource (at least not a remotely up to date one) for such a globally important problem.  As I discussed in Sutton (2019)⁹, whilst the IPCC provides a lot of relevant information, it does not deliver a risk assessment because it is not designed to do so.   As I also argued in the same paper, I believe there is considerable room to improve the process for scoping the major IPCC Assessment cycles to ensure the reports are more sharply focused on the evidence relevant to policy. Of course, AI/ML developments may revolutionise the whole process of literature review and evidence synthesis in ways that can’t yet be fully anticipated.   

The Evolving Role of the Climate Scientist

Lastly, I would like to return to the question of what is the future for climate science and climate scientists?  It should be clear from the above that I expect the demand for climate scientists to only grow.  What will they be doing?  In the first place, they will be central to designing, evaluating and improving the core observational, modelling and exploitation capabilities required to develop a future VIREIS and to deliver to the requirements 1-5 I stated above.  On all these tasks they will be benefiting from powerful AI assistants to optimise designs, interpret evidence, and respond rapidly to unexpected events. Many climate scientists – likely the majority – will be experts in working across disciplines, because they will be directly involved in designing and evaluating solutions to climate challenges and delivering climate services across society. What about fundamental research? The good news for those who are stimulated especially by scientific curiosity, is that the need for fundamental research will endure. The fact is, with the industrial revolution, we began an experiment to perturb an enormously complex system of which we have a very incomplete understanding. So I fully expect the climate system to throw up surprises in the decades ahead, and these surprises will create some of the opportunities for curious climate scientists to further advance human understanding, to the benefit of all.

References

1. IPCC, 2022: Summary for Policymakers. In: Climate Change 2022: Impacts, Adaptation, and Vulnerability. Contribution of Working Group II to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change [H.-O. Pörtner, D.C. Roberts, E.S. Poloczanska, K. Mintenbeck, M. Tignor, A. Alegría, M. Craig, S. Langsdorf, S. Löschke, V. Möller, A. Okem (eds.)]. Cambridge University Press, Cambridge, UK and New York, NY, USA, pp. 3–33. DOI: https://doi.org/10.1017/9781009325844.001
2. My Climate Risk WCRP Lighthouse Activity
3. Chris Kent, Adam A. Scaife, Nick J. Dunstone, Doug Smith, Steven C. Hardiman, Tom Dunstan, Oliver Watt-Meyer,”Skilful global seasonal predictions from a machine learning weather model trained on reanalysis data”, in press, npj Climate and Atmospheric Science, 2025 🔗 https://arxiv.org/abs/2503.23953
4. Kendon, E. J., Addison, H., Doury, A., Somot, S., Watson, P. A. G., Booth, B. B. B., Coppola, E., Gutiérrez, J. M., Murphy, J., & Scullion, C. (2025). Potential for machine learning emulators to augment regional climate simulations in provision of local climate change information. Bulletin of the American Meteorological Society, 2024 
   🔗 DOI: 10.1175/BAMS-D-24-0114.  
5. Peter Van Katwyk, Baylor Fox-Kemper, Helene Hewitt, Karianne J. Bergen, Rewiring climate modeling with machine learning emulators, shortly to be submitted, 2025
6. Explaining and Predicting Earth System Change WCRP Lighthouse activity; Findell, K. L., Sutton, R., et al (2023) A Call to Action: Developing the Capability to Explain and Predict Earth System Change. Bulletin of the American Meteorological Society, 104(7), 501-504. https://doi.org/10.1175/BAMS-D-21-0280.A
7. Safe Landing Climates WCRP Lighthouse Activity
8. Research to Inform Decisions about Climate Intervention WCRP Lighthouse Activity
9. Sutton, BAMS, 2019: Climate Science needs to take Risk Assessment much more seriously, https://journals.ametsoc.org/view/journals/bams/100/9/bams-d-18-0280.1.xml