The Need to Understand Mechanisms: Example, Keeley and Syphard (2025) Paper on Wildfire Drivers

The theme of the week is two-fold. We’ll continuing looking at mechanisms behind correlations, and also how the choice of space and time for studies influences the results.

An excellent example of both is this open-access journal article by wildfire scientists Jon Keeley of the USGS and Alexandra Syphard of San Diego State University. It’s worth a deep dive.

Recent studies conducted across western North America have reported a significant relationship between VPD and area burned and predict that fire activity will increase in the future due to global warming’s intensifying effect on atmospheric aridity (Seager et al. 2015; Williams et al. 2015; Abatzoglou and Williams 2016; He et al. 2025). One limitation of broad-scale studies, however, is that they potentially confound the effects of space and time, as climates and vegetation are variable across western North America.

This may be a version of what I would call “averages are not necessarily true anywhere, so the utility of an average depends on the scale over which you are averaging compared to the scale that you are interested in for doing something.” Some climate policies even notably deal with global average temperatures.. I’ll leave that there. I bolded some key points (to me).

As a result, subregional differences in climate-fire relationships may be masked as noise in aggregated analyses, potentially obscuring areas that diverge meaningfully from the dominant regional pattern. In addition, a major limitation of broad scale studies focused solely on fuel moisture is that they do not account for other drivers of fire regimes such as fuel load, weather, ignitions and management, which vary across broad regions and may benefit from the application of more fine-scale spatial analysis to parse out the relative importance of different drivers (MacDonald et al 2023).

Most folks, I think, would agree that predicting future wildfires without considering suppression (or management) is rather .. silly. But it would be difficult, plus technologies and policies are changing.  So what indeed is the point? And couldn’t tax dollars be used for more realistic research?

In addition to evaluating the relative influence of multiple factors across diverse regions, much of the literature implicating climate change as the causal agent behind increasing wildfires relies on correlations with factors presumed to influence fire behavior. For example, correlations between climate variables and fire activity have been used to suggest that global warming will lead to greater annual extremes in fire behavior (Zhou et al. 2023). However, these studies often lack a causal mechanism (sensu Grace et al. 2025) connecting climate change to fire. For instance, in response to the unusually extreme winds that drove the recent Los Angeles (LA) wildfires in January of 2025, some media reports claimed that the extreme behavior of these winds was consistent with expectations from global warming, and thus attributed the disaster to anthropogenic climate change (Hossenfelder 2025; Qiu et al. 2025). However, establishing causality requires a plausible and testable mechanism. Santa Ana wind speed, widely considered to be the primary factor responsible for the damage from these LA fires (Murphy and Mass 2025), is influenced by the spatial proximity of high and low pressure systems. To date, no model has demonstrated that anthropogenic climate change is altering the distribution of high and low pressure cells in a way that would explain increased wind speed. Indeed, Meehl et al. (2025) reported on an example of foehn winds that suggested recent climate change had contributed to decreased wind speeds relative to historical patterns, and concluded that the mechanism behind this was in need of research. Guzman-Morales and Gershunov (2019) projected decreases in frequency and intensity of future Santa Ana Winds.

When mechanisms are proposed they need to be supported with known causal relationships. For example, Swain et al. (2025) reported correlations that implied high winter/spring rainfall followed by a dry year resulted in high fuel production and this contributed to the catastrophic impact of the January
2025 LA fires. They attributed this to the effect of high rainfall on herbaceous fuel production, a well known phenomenon and something we demonstrate below. However, in this Mediterranean climate herbaceous fuels are always dry and flammable in autumn regardless of whether or not there has been a long-term drought. In addition, as they acknowledge in their introduction, these fires started in chaparral, an evergreen shrubland with little herbaceous vegetation. A high rainfall year may increase primary growth in these shrubs, but these fine fuels comprise a minor part of the total fuel volume (Kummerow et al. 1981), and fire spread is largely a function of the long term accumulation of dead fuels (Keeley et al. 2022). Swain et al. (2025) coined the term ‘whiplash effect’ to describe the sequence of high rainfall followed by dry years (something not unique in the history of this region) and speculated that climate change was going to increase this so-called whiplash effect. However, it is unlikely such an effect can explain the catastrophic LA fires as fire spread was due to urban fuels, primarily the homes themselves, and wildland fuels played little if any role (Witze 2025).

So Keeley and Syphard divided up California into different regions, the North Coast:

This climate division holds the record for the largest fire in the state; two-thirds of the area burned within the August Complex fire perimeter had not recorded a fire since before 1910, conditions suggesting unusually high fuel accumulation. This is indicative of much of the North Coast as the rate of burning during the prior 100 years gives a fire rotation interval of ~130 years. Thus, woody fuel accumulation is likely a major proximate factor in large fires, resulting from ultimate factors that include fire suppression (Stephens et al. 2018), and mesic summer conditions (Viers 1979; Keeley and Pausas 2025), an example of how a proximate driver may have more than one ultimate driver.

The North Interior:

In summary, today climate plays a major role in determining fire regimes in the North Interior, but long-term woody fuel accumulation cannot be ignored as a contributing factor. Predicting future fire regimes based solely on climate projections may be misleading, since with increasingly more area burned (from wildfires and prescribed burn treatments), reductions in fuel load may lead to less extreme fire sizes given fuel feedbacks (e.g. Abatzoglou et al. 2021; Hanan et al. 2022).

Or in my words, if trees are gone and don’t grow back, or grow back at some rate, for awhile there won’t be enough fuel for (some) extreme wildfires.

Sierra Nevada:

In the higher-elevation forests of the Sierra Nevada, dominated by woody vegetation, area burned is dependent on high spring and summer temperatures that reduce woody fuel moisture – a pattern that held true in historical times as well (Taylor and Beaty 2005). It is noteworthy that in all of the forested ecoregions spring VPD was a significant factor driving annual area burned (Fig. 7), and this supports Westerling’s (2016) contention about the importance of early spring snowmelt in determining annual area burned.
However, it seems inescapable that large fires are also driven by anomalous fuel accumulation. For example, threefourths of the area burned in the largest fire (Creek Fire) had no recorded history of previous fire, and the division’s fire rotation interval is estimated at between 112 and 198 years. In light of the known historical range of variation in fire return intervals of roughly 20 years (Stafford et al. 2022), effective fire suppression has contributed substantially to the fuel load. Thus, ultimate drivers of the fire regime appear to be a combination of both climate and fire suppression.
Historically, prior to the advent of aggressively suppressing all fires, these forests experienced a frequent understory fire regime driven by lightning ignitions (Kilgore and Taylor 1979). Under such a regime, fire activity was likely regulated by the time required to accumulate sufficient woody fuels (Stephens et al. 2007), implying that fuel aridity played a secondary role. Looking ahead, increasing focus on fuel reduction treatments (Stephens et al. 2012) may reduce fire size, which over time may ultimately result in reduced available fuels and thus area burned (Kennedy et al. 2021).

There is also the Central and South Coasts, which are probably of less interest to TSW readers, but you can read for yourselves.  This is interesting.

The ignition sources for some of California’s largest fires have increased in frequency and area burned (Tables 1 and 3), and perhaps the most important of these is powerline ignitions e.g. Dixie and Thomas fires (Supplementary Fig. S1b, e). In the recent past powerline expansion into watersheds of hazardous fuels, in the South Coast, North Coast and North Interior have resulted in some of the largest and most devasting fires (Keeley and Syphard 2019; Rolinski et al. 2019; Troy et al. 2022). These types of fires are typically associated with extreme wind events such as Santa Ana Winds in the south or North Winds in the northern part of the state, which cause failures in electrical lines, and the powerful winds ensure such fires become large. It is critical to recognize that the winds per se are not the problem, as the vast majority of such extreme winds do not result in fires; rather fires are dependent on the coincidence of a human ignition source during a wind event (Keeley et al.
2021; Troy et al. 2022).

This is also true in Colorado. From the conclusion:

Lastly, what we have learned from this analysis and the implications for fire management are summarized in a modified version of Fig. 1 (Fig. 10). Here we link the limiting factors to fire activity to likely ultimate and proximate factors and what they suggest in terms of future fire management responses in California and other fire-prone regions in the western US and worldwide. Effective fire management will require aligning interventions with the proximate and ultimate drivers in each region, rather than assuming a uniform role for climate across all landscapes.

Figure 10 is at the top of the post. If you’ll notice, for grasslands grazing is a way of reducing fuels.  We can observe that readily at least around Colorado. Perhaps cows and ranchers are not quite the enemies that they’ve been portrayed by some (remember “Cattle-free by 93?).  Are cows back?  And of course sheep and goats...