One of the downsides of my investigation of evidence-based environmental decision-making being a hobby, is my real life often gets in the way. This means I am not always able to comment on every interesting paper when it comes out. One such example is the paper that came out in January from the University of Victoria titled Attribution of the Influence of Human-Induced Climate Change on an Extreme Fire Season. The paper has been a topic of intense conversation but very little critique. It is repeatedly cited by activists who have not read it, but feel the conclusions:
that the event’s high fire weather/behavior metrics were made 2–4 times more likely, and that anthropogenic climate change increased the area burned by a factor of 7–11.
help their political narrative. I keep expecting to read a serious challenge of its results because it has a really obvious flaw that essentially eliminates its usefulness in quantifying anything; but I haven’t seen one to date. I am surprised because once you see how it treats confounding variables it is impossible to take its quantification seriously. In the rest of this blog post I will provide an explanation for this statement/
Since it is the basis of this discussion lets explain the concept of a “confounding variable” in research design. The simplest description I’ve found online is this
A confounding variable is an “extra” variable that you didn’t account for. They can ruin an experiment and give you useless results. They can suggest there is correlation when in fact there isn’t. They can even introduce bias. That’s why it’s important to know what one is, and how to avoid getting them into your experiment in the first place.
As a practical example, imagine you were comparing death rates from car accidents between 1964 and the present and your hypothesis was that these deaths were attributable to better, modern engines. Confounding variables might include the fact that modern cars have air-bags, better seat-belts and more survivable designs; all features that were not available in 1960’s automobiles. If you did not find a way to correct for the presence/absence of seat-belts, air bags and design considerations then any person reading the study would instantly recognize that the results of the study were invalid.
So how is this relevant to the UVic forest fire study? Well let’s look at what it compares:
As the CanRCM4 ensemble includes natural and anthropogenic forcings, we use the decade 2011–2020 to represent the current climate and an earlier decade and 1961–1970 to represent an alternative climate with reduced influence of human emissions
So much has changed between 1961 and 2011 that I expected to find a lot of work to deal with all the potential confounding variables. Imagine my surprise when I came across this text deep in the report:
The result is dependent on the regression model being realistic [my emphasis throughout]. Our regression model assumes that nonclimatic variability in the natural log of area burned is stationary in time and does not account for the possible influence of human factors such as changes in forest management or human ignition sources. Humans have long had a direct influence on fire activity (Bowman et al., 2011), and trends in some regions have been strongly impacted by human intervention (Fréjaville & Curt, 2017; Parisien et al., 2016; Turco et al., 2014). Syphard et al. (2017) demonstrated that climate influence on fire activity becomes less important with a strong human presence. We also do not consider directly the impacts of repeated suppression over time, which could result in larger fires, nor do we consider the pine beetle infestation that has affected BC
Stop and re-read that section again. Their hypothesis is that climate change is the driving factor but they didn’t correct their work for any of the critical confounding variables. They simply ignored some of the most of the most important considerations when discussing forest fire size/numbers/intensity. Let’s look at them one at a time.
Obviously the first issue to consider is the Pine beetle infestation. As described by Natural Resources Canada:
Over 18 million hectares of forest were impacted to some degree [by the pine beetles], resulting in a loss of approximately 723 million cubic metres (53%) of the merchantable pine volume by 2012. The epidemic peaked in 2005: total cumulative losses from the outbreak are projected to be 752 million cubic metres (58%) of the merchantable pine volume by 2017,
The pine beetles killed massive swathes of our forests and turned them into dead wood just ready to burn. How can a study comparing fire influence without accounting for the pine beetles? Certainly they cite an American study to justify their decision, but numerous Canadian studies indicate that beetle-killed stands have “higher fire spreading potential” among other considerations. Now if the authors had only missed the pine beetle issue it might have been a minor issue but they also missed changes in forest management.
It is well-understood that BC’s forest management has raised the fire risk in BC. BC has systematically been suppressing broad-leaf trees like aspen and birch that provide natural fire protection so as to make room for more commercially valuable conifer species like pine and Douglas fir. Those species are critical to large stands of trees: they’re less prone to burning, create shade on the forest floor, reduce temperatures and promote more humidity. Current forest management has changed the nature of our forests, this is not a hypothesis it has been a stated policy of our forest management regime for decades. How can a study ignore this consideration? Not only have we changed the forest make-up we have completely modified the fire regime via fire suppression.
After the Slave Lake fire in 2011 the Alberta Government sought advice on the fire situation. The result was the Flat Top Complex Wildfire Review Committee Report which made a number of recommendations and concluded:
Before major wildfire suppression programs, boreal forests historically burned on an average cycle ranging from 50 to 200 years as a result of lightning and human-caused wildfires. Wildfire suppression has significantly reduced the area burned in Alberta’s boreal forests. However, due to reduced wildfire activity, forests of Alberta are aging, which ultimately changes ecosystems and is beginning to increase the risk of large and potentially costly catastrophic wildfires.
Essentially the report acknowledged that fire suppression efforts are making wildfires bigger and more dangerous. While the report was written for Alberta the conclusions are entirely transferable to BC. Humans have interfered with the natural regime of fire in order to protect forests for commercial use and we have now created a situation where bigger, and more numerous, fires are a certainty. But that is not all because of the way we have allowed encroachment into interface zones and provided added access to our forests.
Human Encroachment and Access
Another feature the paper missed is human access issues. For those of us who lived through the 1970’s, one thing I can assure you is that access to the back-country has changed significantly since then. In the 1970’s the resource road network did not exist. Huge portions of the province were essentially inaccessible except via air or on foot. This protected the forests from humans and their tendency to light things on fire or drop sparks from their engines. These days we can get to the back-country much more easily which gives more opportunity for fires. Consider this comment from UBC professor Lori Daniels:
The easiest piece of the puzzle is population. There are simply more of us, in more pockets of the province, which inevitably increases the chance of man-made fires. Varying estimates suggest anywhere between 30 to 50 per cent of the current fires are caused by people.
This result is consistent with the BC Wildfire Service which says that 40% of fires are caused by people. The greater access to the back country has resulted in more area under risk from human impacts.
To conclude, let’s look at the confounding variables that were not considered in this study:
- Pine Beetles
- Forest management
- Fire suppression and
- Human encroachment and development
and yet this paper says it can provide accurate quantification of the increases in forest fire activity between the 1970’s and the 2010’s due to climate change?
Like our hypothetical study that ignored seat-belts, air bags and vehicle design, the confounding variables have to have had an effect on the two signature numbers “2–4 times more likely” and “increased the area burned by a factor of 7–11“. Absent controls for confounding variables any quantification of the effect of climate change alone cannot be taken seriously. Certainly, it is entirely likely that climate change will eventually increase the likelihood of fire and even increase the area burned, but those 2-4 times and factor of 7-11 numbers are simply not credible given what we know about the disclosed confounding variables.