Australia's Bushfires Completely Blasted Through the Models

Today in the journal Nature Climate Change, researchers are publishing a series of articles as a kind of postmortem of the Australian bushfires. The series is both a diagnosis of what happened as flames swept across the continent, and a call to action for researchers the world over: Climate change is a crisis for people, the natural world at large—and for science itself.

In particular, some of the research is making a staggering argument: This season’s bushfires were so catastrophic, they caught modelers off guard—way off guard. The models not only hadn’t predicted that bushfires of this magnitude could happen now, they hadn’t even predicted that bushfires of this magnitude could happen in the next 80 years.

“This is perhaps one of the first really big cases where we've seen the real world do something before we've been able to have the capacity to model it properly,” says climate scientist Benjamin Sanderson of the National Center for Atmospheric Research in Boulder, who cowrote a piece in the Nature Climate Change package. “This event was worse than anything in any of the models at any point in this century. Only one of the models toward the end of the century started producing things of this magnitude.”

A model does its best to accurately simulate how the world works by estimating how changes to dozens or even hundreds of variables could lead to different outcomes. For example, climate models simulate how much warming might result from a certain increase in the amount of CO2 in the atmosphere. Political models use data like polling numbers and historical voter turnout levels to predict a candidate’s chances of winning. Fire models use historical observations to make calculations like: In the past, this much vegetation at this level of dryness has led to this kind of bushfire.

Image may contain: Universe, Space, Astronomy, Outer Space, Planet, Night, Outdoors, Moon, and Nature

The WIRED Guide to Climate Change

The world is getting warmer, the weather is getting worse. Here's everything you need to know about what humans can do to stop wrecking the planet.

Yet as the saying goes: “All models are wrong, but some are useful.” Scientists only have access to so much data, and there’s just no way to fully represent the complexities of the real world. Modeling fire is also exceedingly complex, because there are a galaxy of factors that determine the behavior of a fire. A model has to take into account, for example, how forests might be evolving, when a world with more CO2 in the atmosphere might stimulate more tree growth. It has to account for how plant communities might change; perhaps some species will grow more abundant and others more rare. And you’ve got to account for how drought and rainfall might result in more or less brush to burn—a wet year produces a lot of vegetation, which, when followed by a drought, produces mountains of dried-out fuel.

“Fires are right at the end of a long sequence of models which have to be pieced together to get the right answer,” says Sanderson. While it’s easier to model some of the smaller changes that might affect an ecosystem, it’s harder to model scores of them all together and still produce an accurate result. “We have very comprehensive models of forests, and the way that the trees will respond to a warmer climate,” he adds. “And we have very comprehensive models of the climate, and models of fire tuned to individual regions. We're not at a stage where we can put them all together and have confidence in the result.”

Another problem: Running models this complex requires supercomputers. And that’s not cheap, which means scientists don’t get to keep test-running their models to fine-tune them. “It takes huge amounts of energy and computation to run a single simulation,” says Sanderson. “We backed ourselves into a corner with climate science where our models are so computationally expensive, we can't really afford to run them more than once.”