- Climate
Human-driven climate change largely responsible for last 50 years of worsening fire weather in Western North America, new study shows
Each year, wildfires reduce people’s homes to smoldering rubble and leave our natural environments charred and barren. It’s no surprise that everyone wants to assign blame – but the verdicts always seem to vary. So is there a right answer?
Some blame the matchstick, downed powerline, or whatever else sparked the first flame. Others focus on climate change, forestry management practices, and other factors. But in reality, wildfires spread when multiple conditions are met, not just one.
Despite what sparks the initial flame, other factors – such as fuel availability and atmospheric conditions – are what help fires spread. Scientists recently published new insights on ‘fire weather’ – hot, dry, and windy conditions that help wildfires to start and spread – and its 50-year rising trend in Western North America.
The paper was published on 17 May 2025 in Nature Communications Earth & Environment and found that human-driven climate change is responsible for effectively all of the increase in fire weather over that period[1].
In light of these findings – and the fast-approaching 2025 fire season in North America – Science Feedback interviewed three climate scientists with relevant expertise to evaluate this new paper and its significance, and got in touch with one of the paper’s authors to learn more about their findings.
Main Takeaways:
- Climate change is making North America warmer and drier, creating more favorable conditions for wildfires – this is well-established at a continental scale, but a new paper added strong evidence at a regional scale.
- A new study found that human-driven climate change was the main driver of worsening ‘fire weather’ – hot, dry, and windy conditions that help wildfires start and spread – in Western North America over the last 50 years.
- The contributions from human-driven climate change are given as a range; most of that range suggests human-driven climate change is entirely responsible, but even the lowest end of the range shows it’s mostly responsible (81% or more).
- According to scientists we interviewed, these results may suggest that in the absence of human-driven climate change, natural atmospheric conditions would have decreased fire weather in the region over that 50 year period.
- This new paper adds to a growing body of evidence showing that climate change is worsening wildfires – and the conditions driving them – in Western North America.
Human-driven climate change plays a major role in worsening wildfires in Western North America
The Intergovernmental Panel on Climate Change explains that in North America, climate change has led to warmer and drier conditions that are more favorable to wildfires[2]. So where do the findings from the new paper fit in that bigger picture?
To learn more, Science Feedback reached out to the paper’s authors and verified the robustness of the study with other scientific experts experienced in researching wildfires.
Science Feedback contacted Dr. Park Williams*, Professor at the University of California Los Angeles, who studies how climate change affects the hydrological cycle and extreme events like heat waves, wildfires, and flooding. Regarding the recent paper, Williams explained:
A. Park Williams
Assistant Research Professor, Lamont-Doherty Earth Observatory, Columbia University
“Overall I think it’s a good paper and I think it reinforces a general understanding about the connection between climate change and wildfire in western North America: when and where fuels and ignition sources are abundant, if human-caused climate change causes fuels to become drier, then wildfire activity will increase.”
Science Feedback also gained insights from Dr. Daniel Swain*, Climate Scientist at University of California Agriculture and Natural Resources, who studies the dynamics and impacts of extreme events – such as wildfires – on a warming planet. Swain explained to Science Feedback:
Daniel Swain
Climate Scientist, University of California Agriculture and Natural Resources, Los Angeles
“This new study in Communications Earth and Environment adds to the growing body of research which, collectively, overwhelmingly points to a large influence of climate change on increasingly adverse fire weather conditions. In this case, the study focuses on Western North America (including western Canada and U.S.), which is both a highly fire-prone region and one that has been the subject of many previous peer-reviewed studies on the topic.”
*Note: Williams and Swain were not involved with the recent paper.
The evidence linking climate change and wildfires in Western North America continues to grow[3-6]. But scientists have had trouble determining exactly how much humans and nature are contributing to increasing fire weather at regional scales (like in Western North America).
However, in the recent study, researchers used a method called ‘fingerprinting’ to overcome this challenge. Just as detectives can find literal fingerprints at a crime scene, scientists can detect human and natural ‘fingerprints’ in climate and meteorological data like temperature, precipitation, and relative humidity (factors that can influence fire behavior; see Figure 1).
One of the paper’s authors, Dr. James Renwick, Professor at Victoria University of Wellington, explained to Science Feedback how they applied this in a novel way in their research:
James Renwick
Professor, Victoria University of Wellington
“Wildfires tend to occur in warm, dry conditions, when vegetation (the ‘fuel’) is very flammable. We know that climate change is raising temperatures and making evaporation & evapotranspiration more efficient, so land surfaces and vegetation dry out faster. Hence, the two main contributors to fire danger are both ‘enhanced’ as the climate changes. Our paper doesn’t really add to that understanding as that’s the basis for the study. What it [the paper] does show is that it is possible to detect the effect of climate change, the ’fingerprint’ of human-induced change, over western North America, something that hadn’t been shown before. It’s quite an achievement as there’s plenty of ‘noise’ (random weather variability) at the regional scale that can easily swamp climate change signals/trends.”
As Renwick explains above, looking for these fingerprints at regional scales gets complex; things like natural variability and ‘noiser’ data can obscure the fingerprints. In other words, while fingerprints at a global scale are clear – like an actual fingerprint pressed firmly onto glass – fingerprints at a regional scale can be more challenging to detect, like identifying a fingerprint on glass after several others have been pressed on top of it.
To overcome this challenge, scientists in the recent study used a statistical technique called ‘optimal fingerprinting’, which uses advanced statistical analyses to better distinguish between human and natural influence.
As explained to Science Feedback by Dr. Marco Turco*, Postdoctoral Researcher at Universidad de Murcia and Principle Investigator on the PREDFIRE project (climate monitoring and PREDictions to forecast global FIRE activity):
Marco Turco
Postdoctoral Researcher , Universidad de Murcia
“this study enriches the existing literature (but see also: the studies here[3], here[4], here[5], and here[6]) by providing, for the first time, an anthropogenic fingerprinting analysis at a regional scale for a composite fire-weather index, thereby bridging the gap between traditional global-scale attribution studies and regional fire-danger assessments. It delivers direct evidence of how human emissions are altering baseline fire-weather conditions in western North America—an insight that can enhance both predictive capabilities and the targeting of mitigation strategies.”
*Note: Turco was not involved with the recent paper
Figure 2 shows the modeling results – the methods of which were approved of by all three scientists we interviewed – and the human (anthropogenic or ANT) and natural (NAT) trends.
Note that human-driven climate change increased fire weather, while the natural trend decreased it over most of the study areas in North America. The metric they use is the fire weather index (FWI) – a worldwide index used for estimating fire danger. Combining the results of both models over their shared time period of 1959 to 2014, the researchers found human-driven climate change to be the main driver of worsening fire weather over that period[1].
So, overall, the new study enriches the growing body of evidence linking climate change to worsening wildfires – particularly in Western North America. But the exact results may reveal another key insights, as scientists explain in the section below.
Human-driven climate change was the main contributor to worsening fire weather Western North America over the last 50 years
Digging into the specifics, the recent paper suggests that human-driven climate change contributed 81-188% to the observed increase in fire weather in Western North America for the last 50 years. So what exactly does that mean?
Let’s first address a very reasonable question you may have: how is it possible for something to contribute more than 100% to an observed trend? The reason that’s possible – and even normal for fingerprinting analysis – is that some factors strengthen the trend, while others lessen it.
To give a more intuitive example of this, imagine studying a warming trend. There are things that add heat (adding to the trend) or take away heat (subtract from the trend) – in this way something can contribute over 100% to the heating, because the cooling factors take away the ‘extra’ amount.
You can also think of this concept like swimming against the current in a river. When swimming against a current, the water is moving opposite to the direction you want to go and working against some of your effort. Because of this, your movement forward doesn’t seem to match your intense efforts to do so.
The same applies here – and, in fact, potentially reveals an important insight: natural conditions may have been working ‘in the background’ to partly dampen fire weather over the last 50 years, while human-driven climate change was worsening it. As explained by Turco:
Marco Turco
Postdoctoral Researcher , Universidad de Murcia
“The upper bound (≈188%) means the anthropogenic signal alone is actually nearly twice as large as the observed trend, implying that natural variability has, on average, acted in opposition—dampening the net trend by roughly 9–13%”
Adding more context to this, Dr. Park Williams, Professor at the University of California, Los Angeles, explained to Science Feedback:
A. Park Williams
Assistant Research Professor, Lamont-Doherty Earth Observatory, Columbia University
“The fact that a fairly large number of situations indicate that anthropogenic forcing accounts for more than 100% of the observed trend simply implies that, according to these climate models and the methodology used to identify the climate-change signal, the fire weather would have decreased in the absence of anthropogenic climate change.”
However, Williams also explains:
A. Park Williams
Assistant Research Professor, Lamont-Doherty Earth Observatory, Columbia University
“This is plausible, but it’s also plausible that the models the authors used for this study over-estimate the positive effects of anthropogenic forcing on the FWI. This is not the fault of the authors or the modelers; the differences among models in terms of how anthropogenic forcing affect climate simply indicates that there is uncertainty in the effects of anthropogenic climate change, and the uncertainty grows as you zoom in on a given region or look at variables downstream of temperature.”
But, according to the paper, even on the very lowest end of the range, human-driven climate change would still be responsible for most (81%) of the rise in fire weather in Western North America. However, regarding the part of the range above 100% – which is notably the majority of the range – Swain explains that it demonstrates a key insight about natural variability:
Daniel Swain
Climate Scientist, University of California Agriculture and Natural Resources, Los Angeles
“this means that over the full period of record that it’s possible (i.e., the >100% part of the range given) that human-caused climate change would have otherwise resulted in fire weather increases even larger than those observed but for the fact that natural climate variability temporarily acted in the opposite direction, partially masking them. This emphasizes that natural variability can go both ways–sometimes amplifying, but sometimes also diminishing, anthropogenic effects on shorter time scales–and that we should anticipate both accelerations and slow downs in the rate of increase as our baseline stair-steps upward in a warming world.”
Regardless of some uncertainties in modeling methods* – the results show one thing consistently, as Turco explains:
Marco Turco
Postdoctoral Researcher , Universidad de Murcia
“The confidence intervals (lower/upper bounds) reflect uncertainties in both the model response and the observational trend estimate, but they consistently show that anthropogenic forcing is the dominant driver—and indeed stronger than the raw trend itself—over western North American fire weather.”
*Note: a more detailed review of the paper’s methods by scientists we interviewed can be found in the Scientists’ Feedback section below
To learn more about climate change impacts on wildfires, read our past reviews here, here, and here.
Conclusion
Wildfires spread as a result of many factors, not just one. This is often left out of online discussions when people assign blame for wildfires. Instead, these discussions can use black and white language, suggesting, for example, that the presence of one thing (like arson) somehow means that climate change didn’t play a role in a given wildfire. But those types of oversimplifications entirely neglect reality: wildfires need heat, fuel, and the right conditions to spread. In this article, we looked at the ‘conditions’ part of that equation given the publication of a new paper analyzing fire weather drivers in Western North America.
That paper suggests that human-driven climate change is responsible for 81-188% of the worsening fire weather trend in Western North America. The range itself suggests that – even on the conservative side – humans are almost entirely responsible for that trend. However, most of the range is over 100%, which suggests that not only was human-driven climate change entirely responsible for that trend, but also that natural atmospheric conditions opposed fire weather over that time period, helping to dampen the effects. This paper is one of many others in a growing body of evidence showing that human-driven climate change is worsening wildfires, despite other contributions from other factors.
Scientists’ Feedback
Questions from Science Feedback:
- What is our understanding about the influence of human-driven climate change on wildfires? Do you think this paper brings new insights on that connection?
- What is your opinion of the methods used in this paper – in particular, are you convinced that their modeling methods allow for accurate identification of the anthropogenic and natural fingerprints on fire weather?
- One of the main findings in the paper is that “anthropogenic forcings have contributed 81–188% of the region’s observed increasing linear trend in fire weather over the last 50+ years”. Can you explain what that range (81-188%) means – given that the upper range is greater than 100% – and whether it’s well-supported by the paper or more research is needed?
Daniel Swain
Climate Scientist, University of California Agriculture and Natural Resources, Los Angeles
- “There is now abundant evidence from across a wide range of ecosystems and fire regimes globally that anthropogenic climate change has already increased the frequency and intensity of weather conditions favorable for wildfire ignition and spread, and in particular has caused sharp increases in fire weather extremes in many regions. There is also growing evidence that this shift toward more extreme fire weather conditions has directly caused an increase in actual fire size and severity in many settings, as well, especially in primarily forested ecosystems (though there has historically been less data available historically to access the direct effect on fires themselves, as opposed to inferring somewhat indirectly from increases in fire weather conditions).
This new study in Communications Earth and Environment adds to the growing body of research which, collectively, overwhelmingly points to a large influence of climate change on increasingly adverse fire weather conditions. In this case, the study focuses on Western North America (including western Canada and U.S.), which is both a highly fire-prone region and one that has been the subject of many previous peer-reviewed studies on the topic. The present study takes a somewhat different approach in confirming and quantifying a robust climate change influence by using a climate model large ensemble to more clearly separate the contribution by natural climate variability (which varies over time and can cause fluctuations in temperature and precipitation over decades) and by human-caused climate change (which typically has a growing influence over time). But the overall results appear entirely consistent with prior work in the sense that they confirm that human-caused climate change is responsible for approximately all (i.e., 81-188%) of the observed increase in fire weather conditions across western North America since the 1950s. The range in values comes from various uncertainties inherent to the fingerprinting process, including the fact that across broad regions natural climate variability actually would have resulted in a small decrease in fire weather conditions over this period (hence offsetting the observed human-caused increase, and possibly therefore yielding a >100% attributable increase).
Ultimately, this is interesting work and helps increase confidence in previous conclusions strongly linking climate change and western North American fire weather even further by coming to similar conclusions by using new models, methods, and data. But in the bigger picture, this study is notable less for having completely novel or surprising findings but instead for its consistency (despite using different methods) with previous work–highlighting the fact that there is essentially scientific consensus at this point that climate change has substantially increased fire weather conditions both in specific regions of focus and in general globally. - In general, I think the overall approach used here is reasonable. There are some weak points that could potentially affect certain minor results and perhaps the magnitude of the overall topline conclusion, but not the overall conclusion (i.e., that human-caused climate change is responsible for ~100% of the observed fire weather increase in WNA). These might include, for example, not using a coupled ocean-atmosphere modeling framework (which will not capture some key aspects of natural and anthropogenic changes in hydroclimate related to ENSO, for example), or not directly incorporating humidity changes not directly encompassed by relative humidity (i.e., vapor pressure deficit, a variable known to be even more predictive of fire behavior than relative humidity, is not considered here yet has exhibited even more robust observed and predicted changes), or using only two specific climate model ensembles (as opposed to additional large ensembles, which may represent both natural and forced climate variability differently). However, I would emphasize that these are pretty technical reflections that I would be inclined to offer in the context of peer review, rather than something I would discuss in a public setting, as they are very unlikely to affect the topline results or key points (i.e., these are not major problems, but rather reasonable caveats).
- From above: The range in values comes from various uncertainties inherent to the fingerprinting process, including the fact that across broad regions natural climate variability actually would have resulted in a small decrease in fire weather conditions over this period (hence offsetting the observed human-caused increase, and possibly therefore yielding a >100% attributable increase).
Additionally, I would add that this means that over the full period of record that it’s possible (i.e., the >100% part of the range given) that human-caused climate change would have otherwise resulted in fire weather increases even larger than those observed but for the fact that natural climate variability temporarily acted in the opposite direction, partially masking them. This emphasizes that natural variability can go both ways–sometimes amplifying, but sometimes also diminishing, anthropogenic effects on shorter time scales–and that we should anticipate both accelerations and slow downs in the rate of increase as our baseline stair-steps upward in a warming world.”
Marco Turco
Postdoctoral Researcher , Universidad de Murcia
“Opinion on methods
Yes—the fingerprinting approach they use is a well‐established, standard method in the attribution literature. By comparing model runs with and without anthropogenic forcings, they can tease apart the human versus natural contributions to the fire‐weather index trend. This is the same kind of two‐signal optimal‐fingerprinting framework you see in global temperature and precipitation attribution studies, just applied regionally to a composite fire‐weather index. The fact that it passed rigorous peer review and was published in Nature—and that the entire peer‐review file is publicly available—speaks to the robustness and transparency of their methodology. I didn’t spot any critical flaws in their experimental design or statistical treatment.
On values above 100%.
‘Over the maximum shared time window between the models, 1959–2014, the relative contribution of anthropogenic forcings to the observed linear increases in May–October FWI across the region is between ~81–188% as estimated by both models. The anthropogenic contribution counteracts an overall decreasing contribution from the natural response of between –31 to –13% over this period.’[1]
In other words:
– The lower bound (≈81%) means that at minimum, human-driven climate change accounts for about 81 % of the observed upward trend in the fire-weather index.
– The upper bound (≈188%) means the anthropogenic signal alone is actually nearly twice as large as the observed trend, implying that natural variability has, on average, acted in opposition—dampening the net trend by roughly 9–13% (see the NAT contribution of –10.9% to –14.9%) .
Ranges above 100% are routine in optimal‐fingerprinting studies whenever the forced signal exceeds the observed change and internal variability partially counteracts it. The confidence intervals (lower/upper bounds) reflect uncertainties in both the model response and the observational trend estimate, but they consistently show that anthropogenic forcing is the dominant driver—and indeed stronger than the raw trend itself—over western North American fire weather.
Finally, this study enriches the existing literature (but see also: the studies here[3], here[4], here[5], and here[6]) by providing, for the first time, an anthropogenic fingerprinting analysis at a regional scale for a composite fire-weather index, thereby bridging the gap between traditional global-scale attribution studies and regional fire-danger assessments. It delivers direct evidence of how human emissions are altering baseline fire-weather conditions in western North America—an insight that can enhance both predictive capabilities and the targeting of mitigation strategies.”
A. Park Williams
Assistant Research Professor, Lamont-Doherty Earth Observatory, Columbia University
“Overall I think it’s a good paper and I think it reinforces a general understanding about the connection between climate change and wildfire in western North America: when and where fuels and ignition sources are abundant, if human-caused climate change causes fuels to become drier, then wildfire activity will increase.
The main fingerprint of human-caused climate change on wildfires in western North America is warming. Climate modeling experiments like this one applied at the global level have robustly showed that the observed warming trends globally are attributable to anthropogenic forcing for decades, so it’s not a surprise to see that the models they run in this study show that most or all of the warming in western North America is due to anthropogenic forcing.
The warming trends then force a decline in relative humidity, so the finding of an anthropogenic forcing of reduced relative humidities is also expected. Climate models also generally simulate reductions in summer precipitation in the mid-latitudes, though with more uncertainty, so the finding of a forced reduction of fire-season precipitation is unsurprising. While their modeling approach has the weaknesses of using just two atmospheric models that are uncoupled to the land and ocean, strengths are that they have relatively high spatial resolution and large ensembles of >20 simulations were evaluated per model. The large ensembles of simulations run under observed greenhouse-gas and aerosol concentrations allow for characterization of not only the long-term trends caused by the changing greenhouse forcing but also the range of random variability surrounding that change.
The range of variability among the large ensemble of simulations is what gives the range of anthropogenic contributions, from 81-188%. The fact that a fairly large number of situations indicate that anthropogenic forcing accounts for more than 100% of the observed trend simply implies that, according to these climate models and the methodology used to identify the climate-change signal, the fire weather would have decreased in the absence of anthropogenic climate change. This is plausible, but it’s also plausible that the models the authors used for this study over-estimate the positive effects of anthropogenic forcing on the FWI.
Climate models should never be interpreted as perfect, and examples of this are evident in this study, where the effects on precipitation and relative humidity are of opposite sign in British Columbia between the two models and the effects on winds are also quite different. This is not the fault of the authors or the modelers; the differences among models in terms of how anthropogenic forcing affects climate simply indicates that there is uncertainty in the effects of anthropogenic climate change, and the uncertainty grows as you zoom in on a given region or look at variables downstream of temperature.
Notably, as far as I know the authors of this study did not perform the climate modeling experiment. They simply used the outputs from the climate models, which they accessed online. Their unique contribution here was to apply the optimal fingerprinting exercise, a well-established statistical procedure in the field of climate detection and attribution, in an effort to robustly characterize the anthropogenic and natural contributions of anthropogenic climate change to FWI.”
James Renwick
Professor, Victoria University of Wellington
[NOTE: Renwick was one of the authors of the 2025 Nature Communications Earth & Environment paper. The questions Science Feedback asked Renwick were therefore slightly different, as shown below]
- What is our understanding about the influence of human-driven climate change on wildfires? Do you think this paper brings new insights on that connection?
“Wildfires tend to occur in warm, dry conditions, when vegetation (the ‘fuel’) is very flammable. We know that climate change is raising temperatures and making evaporation & evapotranspiration more efficient, so land surfaces and vegetation dry out faster. Hence, the two main contributors to fire danger are both ‘enhanced’ as the climate changes. Our paper doesn’t really add to that understanding as that’s the basis for the study. What it [the paper] does show is that it is possible to detect the effect of climate change, the ‘fingerprint’ of human-induced change, over western North America, something that hadn’t been shown before. It’s quite an achievement as there’s plenty of ‘noise’ (random weather variability) at the regional scale that can easily swamp climate change signals/trends.”
- Can you explain the benefits of the modelling methods used in the paper, and also any uncertainties should be considered when looking at the results of these methods?
“The benefits relate to my last point in (1). There is a lot of noise so to detect a signal, you need a lot of data. The use of atmosphere-only models means we can have access to large data sets of many model simulations, as running a model of just the atmosphere is much faster that running a full coupled ocean-atmosphere model. Hence the modelling centres whose output we used are able to run many more simulations in a set time. That allows us to probe the statistics of the (simulated) climate more closely and deeply. Another benefit is that coupled ocean-atmosphere variability is explicitly excluded, we are assuming the ocean is essentially passive, helping set the scene for atmospheric variability, but not responding in real-time in the model runs. That helps reduce the noise, making it easier to detect a signal. Of course in the real world, the oceans do respond in real time, so our work is a simplification. However, in the middle latitudes such as those in the western North American region, the atmosphere tends to dominate over the oceans in terms of variability, so we don’t think we’re losing anything vital. In the tropics it’s the other way around, the oceans dominate over the atmosphere, but the observed sequence of tropical ocean variability is included in the model sea surface temperature fields, which the atmosphere model would respond to.
A bit of a long answer, but to summarise, we have a large data set because of the modelling setup, so we can get more precise statistics. But, we exclude some of the natural variability, so that’s an uncertainty. Plus we have looked at only two models. So other models, and large ensembles of coupled ocean-atmosphere simulations (when they become available) may show different behaviour. However, the models we used do a good job of simulating the observed climate and we don’t think the ocean component would add much here.”
- One of the main findings in the paper is that “anthropogenic forcings have contributed 81–188% of the region’s observed increasing linear trend in fire weather over the last 50+ years”. Can you explain what that range (81-188%) means – given that the upper range is greater than 100% – and whether it’s well-supported by the paper or more research is needed?
“Yes, this seems a curious result. The real issue here is the mixing of natural variability and human-induced trends in the climate. Over fifty years we may see natural variations (randomness) that happens to counter the climate change trend, at least in part. This is quite prominent at the decadal scale, with 20-year time scale ups and downs in the observations. The technique we used does its best to separate the two, and we do identify the separate patterns of natural and human-induced change, but we can never be precise or separate things completely cleanly. What the numbers are saying is that human-induced climate change alone could have produced 188% of the observed trend, but natural variations has damped that down, so we have the trends we observe.”
References:
- 1 – Queen et al. (2025) Fingerprint of anthropogenic climate change detected in long-term western North American fire weather trends. Nature Communications Earth and Environment.
- 2 – IPCC (2021) Sixth Assessment Report.
- 3 – Williams et al. (2019) Observed Impacts of Anthropogenic Climate Change on Wildfire in California. Earth’s Future.
- 4 – Abatzoglou and Williams (2016) Impact of anthropogenic climate change on wildfire across western US forests. Proceedings of the National Academy of Sciences (PNAS).
- 5 – Turco et al. (2023) Anthropogenic climate change impacts exacerbate summer forest fires in California. Proceedings of the National Academy of Sciences (PNAS).
- 6 – Zhuang et al. (2021) Quantifying contributions of natural variability and anthropogenic forcings on increased fire weather risk over the western United States. Proceedings of the National Academy of Sciences (PNAS).
- 7 – Pausas and Keeley (2021). Wildfires and global change. Frontiers in Ecology and the Environment.
