top of page

Burned Area Increasingly Explained by Climate Change

By: Chantelle Burton, Seppe Lampe, Douglas I. Kelley, Wim Thiery, Stijn Hantson, Nikos Christidis, Lukas Gudmundsson, Matthew Forrest, Eleanor Burke, Jinfeng Chang, Huilin Huang, Akihiko Ito, Sian Kou-Giesbrecht, Gitta Lasslop, Wei Li, Lars Nieradzik, Fang Li, Yang Chen, James Randerson, Christopher P. O. Reyer & Matthias Mengel


Climate change is increasingly recognized as a key driver of wildfire activity, with hotter and drier conditions leading to more frequent and intense fires. But answering the seemingly simple question, "How much has climate change altered burned area?" is far from straightforward. Fires are complex events, heavily influenced by human actions such as land management, ignition, and suppression. Observing trends is challenging due to uncertainties in fire observations; modeling them is even tougher. Fires are shaped by climate and fixed effects like land use, which has decreased burned area in many regions. Our paper tackles these challenges head-on and, for the first time, provides a global and regional estimate of how much climate change has impacted burned areas—shedding light on a critical, yet elusive, climate question.


Modeling fire activity is no trivial task. First, Dynamic Global Vegetation Models (DGVMs) are required- sophisticated tools that strive to represent everything we see on land. These models take in data about the weather, the atmosphere, CO2 concentrations (which also affects plant growth), and human activity – like land use, and population density - and simulate how vegetation grows, the carbon cycles through the ecosystems, and how the land and people interact with the atmosphere and biosphere. But that’s just the first step. To understand fire activity, these models must be linked to a ‘fire model’, a model that predicts how much vegetation burns based on the same inputs + the vegetation and soil characteristics modeled by these DGVMs.


Fires are modeled by considering three crucial elements: ignition, spread, and suppression.  Ignitions can come from natural sources, like lightning or human activity. Lightning is provided as an input, while human ignitions are estimated based on population. Once a fire starts, its spread is shaped by the amount of fuel (vegetation amount and dryness) to burn and, in some models, wind. Suppression is often influenced by population and socio-economical factors like GDP. And, of course, fires don’t just stop at burning— they reshape the ecosystems and alter the carbon cycle, all of which feed back into the complex simulations of the DGVMs (Figure 1).

The skill of a fire model thus hinges on several factors, from the accuracy of the weather and lightning it receives, the performance of the underlying DGVM, and the fire model’s ability to simulate ignitions, spread, and suppression. To test the model’s performance, we compare its predictions with satellite observations of the burned area - though these observations also come with inaccuracies. This long chain of potential uncertainties means pinpointing why a fire model is not doing a good job in a specific instance is no easy task. As a result, the fire models we have today are far from perfect. Each of them has its strengths and weaknesses; one might excel in tropical grasslands, while another might be better suited at simulating fires in boreal forests.


In this study, we made use of seven different DGVM-fire models. Each model was run twice for the period 1901 to 2019. In the first round of simulations, we fed the models with real-world, observed data on weather, CO2 levels, human activity, land use change, and lightning. These simulations represent the planet as it evolved over the last 120 years, we call these the “factual simulations”. For the second set of simulations, we also gave the same observed human activity and lightning, but with a version of the weather-stripped of the influence of climate change. Along with keeping atmospheric CO2 levels low, this essentially recreates how the weather could have looked without human greenhouse gas emissions and climate change. This set of simulations represents the planet as it could have evolved over the last 120 years without climate change, we call these the “counterfactual simulations”.


Figure 1: Schematic of how a fire model works. First (step 1), a DGVM receives weather, land use and population data and uses this to model vegetation. Then (step 2), this vegetation data, along with the weather, land use and population data are supplied to the fire module, which additionally receives information on natural lightning. The fire module calculates how much of the vegetation burns and returns this to the DGVM (step 3).

By comparing these two sets of simulations, we can estimate how much burning is affected due to climate change according to each fire model. However, not all fire models are equally good in each region. Therefore, we compare the factual simulations of each model to the satellite-based observations burned area. From this, we calculate a relative weight for each model for each region. The closer a model matches the observed burned area, the higher its weight will be, meaning it plays a bigger role in the final regional estimate of climate-driven fire activity.


Our work would not have been possible without the available VSC infrastructure.

The top row of Figure 2 shows climate change's impact on each region between 2003 and 2019. Globally, we find a 16% increase in the burned area due to climate change, with particularly sharp increases in Australia, California, and Siberia. Secondly, we also compare the first 20 years of the counterfactual with the last 20 years of the counterfactual. Since the counterfactual assumes a stable climate, any difference between these two periods reflects changes driven by direct human activity such as land use changes and population changes. When we compare these two periods, we find a general decline in the burned area (middle row of Figure 2), suggesting that land use changes and population growth are driving a decrease in the burned area. Finally, when we compare the first 20 years of the counterfactual to the last 20 years of the factual simulations, we capture the combined effect of climate change and human activity (bottom row of Figure 2). This shows a mixed pattern, where some regions have experienced increased burning and others experiencing less.


Figure 2: Changes in the burned area due to different factors. On the top row, the effect of climate change is visible, which is mostly causing an increase in burned areas. In the middle row the effect of human activity (direct human forcings) is shown, which is predominantly negative. On the bottom row, we can see the combined effect of climate change and human activity on burned areas between the periods 1901-1920 and 2003-2019.

In a world where both climate change and human activity are reshaping landscapes, this study offers the clearest global picture yet of how climate change is altering fire patterns. While we see that land management has reduced fire in many areas, climate change is driving increases in others, leading to a complex, region-specific story. As we face a future with rising temperatures, understanding these trends will be critical for developing strategies to protect ecosystems, mitigate fire risk, and adapt to the challenges ahead.


Our work would not have been possible without the available VSC infrastructure.


 

Read the full article in Nature Climate Change here

bottom of page