For decades, one of the most exciting—and controversial—questions in ecology has been: how much does fire shape tropical tree cover? It’s a question that became prominent for many global modellers with Bond’s 2004 landmark “World Without Fire” study, which suggested that fire kept half the world’s forests from existing. That single idea inspired a generation of fire-enabled vegetation models. Yet, despite its influence, the debate rages on, with everything from local experiments to global studies offering wildly different answers, ranging from ecosystems being so adapted to fire that it has little to no impact on tree cover, to the idea fire causes a run-away feedback loop that effectively locks out the formation of forests. Our team set out to revisit this cornerstone idea. Could we better quantify fire’s impact on tropical ecosystems using modern global remote-sensed data and advances in statistical techniques? 

The world without fire

Imagine a world where fire did not exist. Would the savannas of Africa and South America turn into dense forests? Would tropical ecosystems look fundamentally different?

Bond et al. (2004) argued that they would. The study introduced fire vegetation models to the idea that fire creates a feedback loop in savannas: burning reduces tree cover, promoting fast-growing, flammable grasses, which in turn encourages more fire. This “savanna-fire feedback” seemed to explain the stark contrast between open savannas and closed forests in areas with similar climates.

But analysis from remote sensing, long-term fire experiments, and fire exclusion studies have painted a more complex picture. Some suggest that fire has a limited impact on tree cover, while others show dramatic increases in cover when fire is suppressed. The truth, as it turns out, is somewhere in between—and far more nuanced.

Revisiting Fire's Impact with Modern Tools and Data

To tackle this question, we turned to an advanced statistical technique called Bayesian inference. This method lets us account for uncertainty and capture the messy, nonlinear relationships between fire, climate, human, and tree cover. It’s a significant leap forward from the modelling approaches available two decades ago and combines with data from the satellite observational era that was only just dawning during Bond's 2004 study.

What did we find? Fire does reduce tropical tree cover—but its impact is much smaller than previously thought. Across the tropics, fire accounts for just 0.3–3.2% of missing tree cover extent, rising to 0.3–5.2% when excluding human influences like fire suppression and deforestation. In dry savannas, the numbers are slightly higher—0.6–7.1%, or up to 0.8–10.8% without human impacts.

These results challenge the assumptions built into most fire-vegetation models, including those in the Fire Model Intercomparison Project (FireMIP). Simply put, fire’s impact on tree cover is much less than these models suggest.

A tale of two tropics

So does this mean fire doesn’t matter? Not at all.

Our study highlights a critical distinction between ecosystems. In tropical savannas, trees have evolved to tolerate fire—thick bark, resprouting, and fire-triggered germination help them recover quickly. This resilience is the likely reason that fire has less impact on tree cover than previously believed.

But in tropical forests, where trees lack these adaptations, even small increases in burning can cause significant tree cover loss. For example, we found that a 1% increase in burnt area could lead to a 2% reduction in tree cover in some forested regions, such as the eastern Amazon and Indonesia. These forests are already seeing more fire due to climate change and human activity, making them highly vulnerable.

Our study also looked at average conditions between 2000 and 2013, providing a snapshot rather than a dynamic view. While fire plays a role in shifts between forest and savanna states, our analysis didn’t explore these temporal dynamics in detail. Other factors, like deforestation or drought, often interact with fire in complex ways.

Bridging Local and Global Perspectives

One of the most intriguing findings was the disconnect between local fire experiments and global models. Fire exclusion studies in savannas often show large increases in tree cover when fire is removed, seemingly supporting strong fire impacts. Yet these local-scale results don’t translate to the broader patterns we see across the tropics.

Why the discrepancy? One reason could lie in how models distribute fire. Most assume that burnt areas affect entire grid cells—often spanning thousands of square kilometers—equally. In reality, fire can be highly localized, in some places repeatedly hitting flammable grasslands while leaving forested areas untouched.

Another factor is the lack of fire-adaptive traits in models. Without accounting for the diverse ways trees respond to fire, models can’t capture the full dynamics of fire-prone ecosystems.

Why this matters for carbon

These findings have big implications for carbon storage.  Tropical forests and savannas are vital carbon sinks, and fire plays a complex role in shaping their carbon dynamics.

Our findings suggest that current fire-vegetation models overstate fire's impact on tree cover , and therefore carbon fluxes, in savanna ecosystems—areas where fire and vegetation are often in equilibrium. In these regions, models should focus less on fire itself and more on non-fire-related factors to explain tree cover dynamics, such as rainfall variability or soil properties.

However, the carbon impacts of changing fire regimes—alterations in fire frequency, intensity, or timing—may be far more significant in tropical forests. These ecosystems are not fire-adapted, meaning even small shifts in fire activity can lead to substantial tree loss and carbon release. This distinction underscores the need for models that capture the different ways fire influences forests and savannas under changing climate conditions.

The path forward

Our study points to several areas for improvement in fire-vegetation models:

• Sub-grid heterogeneity: Representing the patchy nature of fire within model grid cells.
• Fire-adaptive traits: Incorporating traits like thick bark, resprouting and shifts in canopy height and structure to better simulate tree responses to fire.
• Emerging fire regimes: Shifting the focus from absolute fire impacts to how changing fire patterns affect ecosystems, especially in vulnerable ecosystems. Some models are beginning to capture these changes, even if estimates of absolute impacts remain imperfect.

Ultimately, this work is not just about refining models. It’s about understanding how fire shapes our world—and using that knowledge to protect the ecosystems that sustain us. As climate change accelerates, the way fire interacts with ecosystems is shifting in unprecedented ways. Recent years have seen fire extremes devastate regions like the Amazon, Australia, and California—events that are reshaping ecosystems and threatening their ability to store carbon. Understanding these dynamics isn’t just an academic exercise; it’s critical for mitigating future risks.

By refining our models and focusing on how emerging fire regimes will impact forests and savannas differently, we can better predict the consequences of these changes. This knowledge is essential for guiding our attempts to protect biodiversity, sustain global carbon sinks, and improve adaptive management strategies in a world where every fraction of a degree of warming counts. Fire has shaped our planet for millions of years, and our ability to manage it will help shape the future of life on Earth.