ľ�ϸ���Ӱ��

To explore what lies ahead, scientists rely on climate models. These models act as virtual versions of our planet, showing how temperature, rainfall, sea level, and extreme weather may change under different emission scenarios. But just how reliable are they? Utrecht climate scientists Roderik van de Wal and Michiel Baatsen explain both the strengths and the uncertainties of these powerful tools.

We would all love to know what the future holds, whether in ten, fifty, or even a hundred years. While we cannot see ahead with complete certainty, climate models bring us remarkably close. These models function like digital versions of Earth, allowing scientists to explore how temperature, rainfall, sea level, and extreme weather might change under different levels of greenhouse gas emissions.

Climate models do not offer precise predictions; instead, they reveal long-term trends under specific scenarios. For instance, a model may project that average temperatures in some country will rise by two degrees by 2050 if high emissions continue. This does not mean every day will be warmer, cold days and regional variations will still occur, but the overall trend points to a steadily warming climate.

Greatest source of uncertainty

Climate models make one thing crystal clear: the Earth is warming, and human activity is the main cause. On this point, most climate scientists are in full agreement. The uncertainties lie not in whether the climate is changing, but in exactly how it will unfold. How quickly will global temperatures rise? What roles will complex systems such as ocean currents and ice sheets play?

Above all, the biggest unknown is us. The greatest source of uncertainty over the medium and long term is the scenario chosen for the simulations, because these depend on future human choices about emissions.

The physical foundation of climate models is robust

Robust

One cornerstone of climate models is beyond dispute, says Roderik van de Wal: the Earth’s energy balance, which determines whether the planet warms or cools. “Because of rising carbon dioxide emissions, more and more energy is trapped in the system. We know for certain that this is driving global warming,” he explains.

Assistant professor Michiel Baatsen agrees: “The physical foundation of the models is robust. The main uncertainties lie in complex, dynamic processes such as ocean currents—how exactly they affect the climate, and over what timescales.”

Melting of the ice caps

The melting of the ice caps is a striking example of a complex, dynamic process with far-reaching effects on the climate. It not only raises sea levels but also disrupts ocean currents and weather systems worldwide. “Melting processes are often calculated separately, while ideally they should be included in the model because of the feedbacks involved,” explains Van de Wal. A feedback occurs when the outcome of a process influences the process itself. In this case, melting reduces the amount of ice, which means less sunlight is reflected. The darker surface absorbs more heat, accelerating the melting even further.

Another key uncertainty is timing. How quickly are the ice caps shrinking? “For climate models it makes a huge difference whether this happens over decades or over centuries to millennia,” notes Baatsen. “The melting of the ice caps represents a tipping point: an irreversible threshold that can trigger major shifts in the climate system.”

Because tipping points carry such profound consequences, scientists have intensified their study of them in recent years. As a result, climate models are becoming better at showing how such thresholds arise and unfold. However, the uncertain timing of these events also exposes new uncertainties within the models. 

Estimates within the model

Another major source of uncertainty in climate models lies in the use of simplified formulas to represent small-scale but complex processes, such as cloud formation and turbulence. These processes strongly influence the climate, but scientists are not yet able to simulate them with full accuracy. Instead, they rely on simplified descriptions known as parameterisations. Baatsen points out that cloud physics in particular remains highly parameterised, which creates uncertainties, especially when environmental conditions change.

Models become more reliable when parameterisations can be replaced with actual physics. Achieving this, however, requires vast amounts of data and a deeper understanding of the underlying processes. First, extensive measurements are needed to capture exactly what happens in reality. Then those observations must be integrated into the model. “We are steadily adding more components and their interactions to climate models, which means the physical core of the models continues to expand,” says Baatsen.

But with greater complexity comes a demand for more computing power. Today’s supercomputers are already pushed to their limits, and the available computing capacity determines how detailed and sophisticated a model can be. Progress in computing technology is therefore just as crucial as scientific advances in making climate models more accurate.

Our models describe temperature and precipitation patterns of the past hundred years very precisely, even on a regional scale

External factors

Climate models carry uncertainties not only because of their own limitations but also due to external factors . For instance, we cannot know exactly how much greenhouse gas humanity will emit in the future or how land use will evolve. To address this, scientists rely on different scenarios, though which scenario will ultimately prove most realistic will only become clear over time.

Another challenge lies in the nature of the climate system itself. It is not deterministic but chaotic: long-term trends can be projected, but precise conditions on a specific day cannot. “Even if we could fully describe the climate system with physical laws, it would still be impossible to predict the future with complete certainty,” explains Baatsen. “Tiny changes can have major consequences—for example, the sudden release of methane from thawing permafrost. This inherent unpredictability is part of the system.”

Predicting the past

Despite the uncertainties, scientists do not take climate model predictions at face value. They rigorously test their reliability—by comparing different models, matching outcomes against real-world observations, and even using them to reconstruct climates from the distant past, sometimes going back fifty million years. These are known as palaeoclimate simulations. “Conditions were very different back then,” says Baatsen. “If our models can successfully reproduce ancient climates, we can trust them to simulate future changes as well.”

According to Van de Wal, however, you don’t need to look that far back to assess a model’s accuracy. A century is often enough: “Our models describe temperature and precipitation patterns of the past hundred years very precisely, even on a regional scale. That gives us confidence that they can also capture the climate of the next hundred years.”

Prepared for the future

Climate models may not deliver exact predictions, even with the best data and the most powerful computers, but they do offer a reliable picture of how the climate is likely to evolve. Their physical foundation is solid, and it grows stronger as scientists gather more data and refine their understanding. This steady progress makes models increasingly precise, allowing them to project expectations for smaller regions and shorter timescales.

By showing how the climate is likely to change under different scenarios, models reveal the risks we face and highlight how much emissions must be reduced to limit those risks. In this way, they are an indispensable tool for preparing for the future.