Categories
Can you spot the problem with this headline?
Here’s a hint: It contains a common misconception about how scientists use mathematical models, one that sometimes occurs even among scientists themselves.
The answer? The Intergovernmental Panel on Climate Change (IPCC) is not predicting, but projecting possible catastrophe. This seemingly minor conceptual error can lead to serious misunderstandings of what model research tells us.
To clear up the confusion, we first need some background on what models are and how they work.
Two kinds of models
“A model, at its most general, is a representation of reality—a simplification of reality,” said FEWscapes researcher Monica Turner, professor of integrative biology at the University of Wisconsin-Madison.
Some models are purely conceptual. Others might be physical, Turner said, like miniature buildings architects and engineers create before breaking ground. But in the scientific world where FEWscapes is situated, we often use mathematical models, in which equations are designed to represent more complex real-world dynamics and then solved on a computer.
We can break these mathematical models into two categories: those that aim to explore the future versus those that aim to predict it.
There’s one kind of predictive model that many of us use every day: a weather forecast. It uses weather data to create a simplified representation of the infinitely complex interactions that make up our atmosphere.
And from that representation, it issues predictions: The temperature today will range from 50 to 70 degrees. There’s a chance of thunderstorms tomorrow. It will rain this coming Saturday.
A focus on the near future is crucial for most predictive modeling. “Weather models are grabbing real-time data and constantly reevaluating their predictions to increase the accuracy of their forecast,” Turner said.
Contrast weather forecasts with a form of exploratory, non-predictive modeling: long-term projections of future changes in the global climate.
“If we say we want to predict, we want to know what will happen,” said Turner. “That is not possible to do if we’re going out 50 years or 100 years.”
Projection is not the same as prediction
To understand possible events decades into the future, Turner explains scientists use exploratory models to simulate “if-then” scenarios – i.e., if this happens, then what are the logical consequences to be expected under those conditions?
So, climate modeling of the sort done by most climate scientists is a projection rather than a prediction: an expected outcome contingent upona specific scenario or set of model “inputs,” such as humanity’s future greenhouse gas emissions.
The results of exploratory models can easily be confused with predictions, especially if they’re described loosely in media coverage, such as the headline above.
The IPCC is an international body that synthesizes the latest climate change research. It doesn’t predict what will happen to our future climate, but projects what could happen based on variable assumptions.
In other words, it constructs the “if-then” style scenarios described by Turner. If humans emit a given quantity of greenhouse gases, then what changes in the climate can we expect?
This isn’t just a matter of nitpicking about word choice. The framing can have a huge impact on how we understand the role of models in research.
If people think the IPCC predicts disaster by 2040, one possible response could be to conclude there’s nothing we can do to prevent that disaster, and it would be futile to try; we might as well keep burning fossil fuels and enjoy the ride while we still can.
But the IPCC’s models are not saying with certainty that humanity can expect disaster by 2040.
Take a look at the IPCC report referenced in the headline we quoted at the top of this post. It communicated its findings with language like this: “Limiting warming to 1.5°C depends on greenhouse gas (GHG) emissions over the next decades, where lower GHG emissions in 2030 lead to a higher chance of keeping peak warming to 1.5°C (high confidence).”
This language makes clear that the model is a projection contingent on future events – namely, humans’ greenhouse gas emissions. Our actions influence the amount of emissions, and our influence over those future events can change the expected climate outcome.
The amount of emissions is not the only “if” climate modelers have to estimate. If we know how much greenhouse gases humans will put in the atmosphere in the future, then modelers can make a pretty close estimation of how much the planet will warm – but they can’t know how much exactly.
The IPCC addresses this uncertainty about how much the planet might warm through its probabilistic language (lower emissions mean a “higher chance” of limiting warming), as well as its “high confidence” parenthetical, both of which acknowledge the challenge of modeling a subject as complex as the world’s climate.
The confusion of prediction versus projection is a common trap in news accounts of climate models. In fact, a 2024 study in the journal Climatic Change found that around 20% of the English-language media coverage analyzed used prediction-based language.
(And in fairness to the website we reference above, the content of their story captures these dynamics much more clearly than their headline.)
Scenarios for the future of the Upper Mississippi River Basin
The FEWscapes project will soon be releasing the results from modeling four exploratory scenarios that project possible futures of the Upper Mississippi River Basin in 2050.
These scenarios follow the “if-then” construction described above: Ifa specific set of changes occur on farms, on the landscape, and in the global climate, then what results can we expect in terms of water quality, food availability, bioenergy production, and ecosystem health?
What might happen if we implement conservation practices on most farms? Convert large swaths of farmland to grazed prairie? Do widespread restoration of forests and wetlands? Or a mixture of all of those things?
We’re using a few mathematical models to explore these questions: process-based crop models, which simulate biological, chemical, and physical dynamics that influence plant growth; water routing models, which represent the movement of water and nutrients through a landscape; and general circulation models, which simulate the earth’s climate system under present and future conditions.
The scenarios are based largely on ideas from regional stakeholders about how to improve food, energy, water, and ecosystem security. Neither the scenarios themselves nor the modeled impacts of those scenarios will be predictions of the future.
What, then, is the use of models that don’t actually tell us what is going to happen? Why use a tool that comes with all the uncertainty and caveats described in this post?
The answer is, above all, they use real-world field data to offer us a window into possible futures that we couldn’t otherwise glimpse. And these possible futures can tell us something about what we could do today to achieve desirable outcomes.
For one, field data alone capture particular moments in time but cannot tell us what will play out in the future.
“With my field data, I can’t tell you what will happen in 50 years, or I can’t tell you what will happen if the weather is different than what I’ve seen over my five years of studying,” said Turner.
Field experiments are also constrained by real-world feasibility. In the case of FEWscapes, it would be impossible to actually convert a large percentage of agricultural land in the Upper Mississippi River Basin, just to see what happens.
But in the simplified, simulated world of the model, it is possible to experiment with landscape changes. The results can help us explore the range plausible future possibilities in this landscape.
If farmers transition to perennial agricultural systems like managed grazing, what changes could we expect to water quality and food production, especially in a climate warmer than what we’ve historically seen?
If our number one goal on the landscape is to improve water quality, what land uses are more likely to bring us closer to that goal: managed grazing or reforestation? Cover cropping corn acres or converting them to restored wetlands?
We also hope our experiments on the simulated model landscape will help stakeholders think through tough tradeoffs. For example, if we want to prioritize the creation of more wildlife habitat in the region, how much corn and soy production would we have to give up?
We hope the FEWscapes model results will help decisionmakers formulate strategies to reach food, energy, water, and ecosystem goals based on an improved understanding of possible future conditions. We’re in the final stages of crunching the numbers now, so stay tuned for more from us soon.