Introduction
I wrote a post on wildfire shortly after I retired that was, I think, an objective account of my experience working as a biologist involved with fire effects and post-fire recovery. While I stand by that account, the rhetoric respecting wildfire and climate change has gotten even more heated, so I thought an update might be warranted. The stimulus for this was, interestingly enough, an update (or perhaps completion is a better word) to the NIFC fire history website. The update consisted of extending the record back to 1926, several years earlier than the website included when I penned the earlier post. (Also see update posted below.) While obviously more complete, the updated history contained a disclaimer about data before 1983 similar to the one that the earlier page contained:
Prior to 1983, sources of these figures are not known, or cannot be confirmed, and were not derived from the current situation reporting process. As a result the figures prior to 1983 should not be compared to later data.
When I saw the update, I decided to update my graphics to match the extended record. I was anxious to do so because I knew older data existed because the U.S. Forest Service had published it as part of their Forest Health Protection monitoring (more on these data below). As previously. I created two graphs, one showing extent (i.e., acres burned) and numbers of starts (i.e., ignitions):
Historical and Ecological Context
It is, of course, a well known fact that climate affects wildfires and that wildfires have been a potent driver of ecological change. The scientific literature is quite clear that climate changes in the paleo-record show clear changes in natural vegetation that can be described by studies of shifts in dominant pollen types, charcoal left behind by wildfires, etc.; some of these are provided in my previous post. The first year that I worked on the Payette National Forest was 1994 and marked probably the worst fire season that had occurred here, the Forest that actually developed "smoke jumping" and serves as an important national fire resource center, with about 300,000 acres ultimately involved. These fires followed several years of sustained drought over much of the Forest (see Johnson Creek discharge data and PDSI plot below***), and it seemed obvious to me that this was, indeed, the beginning of a period of climate-forced ecological change; the fact that these fires burned in large swaths of spruce and fir forests that are not fire-adapted species (e.g., they are readily killed by fire and have no specific adaptations for rapid regeneration, flame avoidance, heat tolerance, etc.) that contained trees that were potentially as old as 400 years heightened this belief. Since that time, nearly a million more acres of the Forest have burned in several large fire events, but the link to climate change has, in my mind has weakened for this area. Part of the cause of my diminished belief in the role of climate here is that the fires, while occurring in relatively low water yield years, have not been associated either with extensive drought (detailed information here) or any perceptible change in temperatures.
It seems that much of the general public either don't understand the prevalence of fire in the western United States or are simply pushing a political agenda that includes "climate hysteria." The year 1994 was my first experience with wildfire as a professional, and it seemed disastrous at the time: I'd finally moved up to the career posting of my dreams and now the Forest seemed to be literally going up in flames. I was conscripted to be the fish resource technical specialist on the post-fire response teams, however, and that experience really broadened my perspective. As an ecologist, I understood in general terms the role fires play in various ecological settings, but had no personal experience before this; so I learned. I learned that the fires, regardless of whether they are unusually large or simply typical, that there are benefits and costs to both ecosystems and humans (some of this is discussed in my previous post, and I hope to write a post focusing on leveraging fire to improve conditions in the future). In the northern Rocky Mountains where I live, the most famous fire season (prior to 2007, at least) was the Great Fire of 1910, which probably comprised more that 3,000,000 acres, but we get some fires, usually lightning-strikes, every year. But going back in time, it is pretty well known that very large fires were not unknown. A Forest Service review indicates that fires as large as 988,000 acres occurred in Pacific Northwest forests in the 1700s and a 432,000 acre fire charred Washington in 1902. Even after fire suppression techniques and record keeping were improving, the Tillamook Burn in Oregon burned 227,000 acres in 1933 with several re-burns over the next two decades.
Sometime in the mid-20th century, people began to become concerned with fire suppression as an unnatural source of ecosystem disturbance that allowed fuels (woody debris, low shrubs, etc.) to accumulate to levels that would not have occurred in areas that naturally experienced recurring low-intensity fire events. The U.S. Forest Service came to embrace this idea and "let it burn" ideas began to appear around 1960 resulting in official policy by 1970†. Here is a graph produced by the Forest Service in the Forest Health document mentioned above showing a rapid decrease in fire acres during the suppression era:
It clearly includes the area NIFC now considers unreliable, but was useful to the Forest Service up to 2010 at least for selling "let it burn" and "fuels management" projects. Here is the language that accompanies the chart to help let this point sink in (emphasis added):
Fire is the most dominant abiotic agent in terms of area affected across the landscape, but is also an integral part of many forested ecosystems. Between 1945 and 2000, fire suppression substantially reduced annual acreage burned. Since 2000, an increase in area burned has occurred, although it has not yet reached the levels recorded between 1925 and 1960.
Out here, most of the state is largely uninhabited, mostly federally owned and managed, with extensive wilderness and relatively pristine forest. Letting fires burn doesn't carry much risk in terms of real or personal property and danger to residents and visitors. Let it burn policies make some sense outside areas designated for timber harvest so long as property and lives are protected. But in 1988, Yellowstone National Park caught fire. Suddenly, even though people think of Yellowstone as "natural," people saw a national treasure completely reconfigured as fire engulfed about 36% of the park (800,000 acres). The decision was made to fight the fire rather than let it burn naturally, and the idea of "let it burn" as a national policy for wildland fire came up for review. But the concept was not abandoned and land management agencies moved back towards tactics based on the fire is an important ecosystem process approach. In the case of the Forest Service, the Nation's largest land management agency, new forest planning rules were adopted in 2000 and revised in 2005, and the "fire use" concept was born. The National Forests of the Southwest Idaho Ecogroup (the Payette, Boise, and Sawtooth National Forests) were the first Forests to revise their Forest Plans under the new rules; here is some key language in these plans:
Appropriate Management Response (AMR)
Actions taken in response to a wildland fire to implement protection and fire use objectives.
fire use
The combination of wildland fire use and prescribed fire application to meet resource objectives.
wildland fire use (for resource benefits)
The management of naturally ignited wildland fires to accomplish specific prestated resource management objectives in predefined geographic areas outlined in Fire Management Plans.
The take away message from the above background discussion is that fire extent ("acres burned") may or may not be related to climate change, but it most certainly is related to fire management strategy. The fires in 1994 that I mentioned above was largely fought with a hands off approach. They burned primarily in forest not designated for logging (including designated wilderness), contained little private property or infrastructure (which were mostly protected), and were allowed to spread where they didn't endanger such things. They grew to over 300,000 acres. In contrast, a fire started behind my house this past summer. It was aggressively fought (see image below). It started at night and, since fires out here are not fought at night, it grew overnight to about 90 acres; about a week later it was effectively extinguished at about 91 acres. Consequently, regardless of whether historic data are "good" or "bad" compared to modern data, I regard extent as a misleading indicator of fire activity. These appear to have either decreased over time, been relatively stable since 1983, or both.
A Different Analysis
When I was thinking about this dilemma about data reliability, I decided to try a different approach. The USFS chart and rationale for the efficacy of fire suppression posted above made me wonder whether I could somehow remove the or reduce the potential error described for the pre-1983 data. It occurred to me that calculating an index based on the ratio of extent to starts might offer a way. This seems reasonable because each fire would be associated with an extent and an ignition, so creating the index would remove the potential problem of counting an individual fire multiple times. It would not correct errors in either the extents or the actual number of starts, but it may well reduce the total error and make the plot across years more accurate. It also suggests two logical hypotheses that can be, if not reliably tested, at least be evaluated graphically to some degree:
Hypothesis 1: If fire suppression (or "the suppression approach") was highly effective, the ratio of extent to starts should decline over time as suppression techniques improve.
Hypothesis 2: Let it burn policies (or "the ecological approach") should cause the ratio to increase relative to the suppression approach.
Before calculating the ratios of extent to starts by year, I divided the data record up into parts. For this exercise, I use these periods:
1926-1959: Extinguish Policies
1960-1999: Mixed Policies
2000-2019: Use Policies
For the graphic, I plotted the size to starts ratio by year, fitting a simple linear model to each strategy, resulting in this graphic (important events are still shown):
The resulting curves seem to comport with the two hypotheses pretty well in my opinion. During the extinguish all fires period, the ratio steadily declined until about 1960, and, although the ratios stabilized a bit after 1960, they still gradually declined until the break at 1983. There is undoubtedly some residual problem with these data because they're provenance is not well known and most records were almost certainly paper; it seems reasonable that the further drop in about 1970 is related to the implementation of infrared mapping in 1969, which was almost certainly more accurate for perimeter mapping than visual methods that probably included aerial and ground-based techniques. I can't readily explain the increased dispersion from 1983 to 2000 except by thinking there is a change from poorer data management and sharing at the 1983 demarcation to better sharing after. There may be a change in accuracy, but it seems likely to me that the most significant thing that happened around 1983 was the installation of a corporate computer network with terminals for individual employees. We called this "the DG," which was short for Data General, the manufacturer of the hardware††, and corporate data sharing and database development was vastly improved. Nonetheless, the greater dispersion and apparently increasing ratio suggest fires gradually getting larger into the 21st century.
Concluding Thoughts
Clearly, the preceding is not a thorough analysis of the effects of climate change on wildfire, but it is not intended to be. The intent was to show that rejecting data prior to 1983 because the NIFC website does not enthusiastically support data prior to 1983 is insufficient reason to disregard it, and the fact that a writer at Carbon Brief talked to (apparently) a Public Information Officer (PIO) at NIFC who didn't know the provenance of the earlier data does not offer much additional support. I have worked with several PIOs in may career, and I would not have expected them to know the details of the data I was managing because that is not their job; they would have repeated what I told them. This is not to denigrate PIOs and the PIO in this case may have been familiar with the underlying data or have conveyed a data manager's views to Carbon Brief, but no such information is provided. In fact, the PIO is quoted by Carbon Brief as saying "for all I know...", a clear suggestion that he was not well briefed on the data management but did have an agency policy response. I think that response is appropriate but not very informative and not sufficient to preclude additional analysis. Furthermore, Carbon Brief was either ignorant of or unwilling to disclose information that weighed on either the quality of any of the data, why the quality may have changed, whether the indicator "acres burned" is actually a reliable metric for associating wildfire extent with climate change, or why, despite their use of the term "increase in wildfires," the actual number of wildfires continues to decline. It does seem likely that even if the early data are less reliable than newer data, the general decline in burned area during the era of vigorous suppression occurred, just as it's obvious that really large fires are associated with dry years. If their intent was to demonstrate that climate change is driving more and larger fires by excluding older data and using fire extent in recent years only, I think they have failed to make a solid case and have provided my case for why it is overly simplistic and analytically flawed, if not deliberately misleading.
This discussion would be incomplete without considering the administrative incentives involved. I hinted above that PIOs are paid to pass on a narrative in response to inquiries, and I say that as someone who has been involved with releasing public information, not as someone who disrespects them or their offices. But if you look at the Forest Health Protection chart above and read the accompanying text, you can see an example in, particularly, the bolded text. This was the era of getting the public to understand that policies such as fire use and controlled burning were essential management tools. It indicates that fire suppression was so good that fuels were unnaturally heavy so these projects and practices were ecologically necessary. That is almost certainly true for some forest types, but there is also research suggesting that these practices may be counterproductive in other forest types. Along with this came the concept of "uncharacteristic wildfire" suggesting that intense or severe wildfires were inevitably unnatural because of decades of fire suppression and climate change even though research in the same areas may suggest otherwise.So now we need more fuels projects not only to restore the forest but also to minimize the threat of "uncharacteristic wildfire." In the meantime, firefighting expenses rise while efficacy seems to decline, and fuels reduction is used to justify timber harvest in areas where it should be avoided (like residual riparian areas in over-harvested watersheds) while wilderness areas burn to the ground under fire use management.
The last few items represent personal complaints during about 20 years working with fires, fire research, fire staff, and timber management. But the initial point remains: The attribution of modern wildfire extents to climate change are overly simplistic and reflexive use of it represents a deep misunderstanding of the issues. In particular, drought, as represented by water yields in this area, has essentially no relationship to atmospheric CO2.
Update 03-21-2021: The NIFC Fire Statistics page has been changed since this was first posted and no longer contains pre-1983 data. This is not surprising because the older data do not reflect the current agency mission as described above. My experience has been that the USFS re-interprets history to suit current management objectives and cannot be trusted without question. I have been unable to resurrect the older pages from the Web Archive at this point, but it appears that the link I provided above redirects to a new URL. The original URL was:
https://www.nifc.gov/fireInfo/fireInfo_stats_totalFires.html
and not as it is now:
https://www.nifc.gov/fire-information/statistics/wildfires
This may help anyone who wants to search the Web Archive for themselves. Tony Heller has also recently posted on this issue.
__________
*The graphics have been updated with aesthetic changes consisting mostly of improved annotations since originally posted, and the charts on initial posts differ from the above.
** The climate alarmist website Carbon Brief made a big deal out of this in 2018, even going so far as to interview a NIFC employee about the reliability of the older records, which, unfortunately in my opinion, is also reiterated at Our World in Data where they assume the disclaimer also means that all post-1983 data records are comparable without detailed explanation.
***I shaded the period from 1981 to 1995 to show how low the river flow was during the period right before I joined the Payette National Forest, with the red dashed line showing a declining trend. This is not intended to match up exactly with the positive PDSI period in the following chart, but they obviously overlap.
†The term "let it burn" is not popular with fire managers because even fires where suppression strategies are not implemented are still monitored, but monitoring is not suppression no matter how much one spins it.
†† It's easy for people to forget that the computer tools they take for granted now were pretty primitive even in 1983. The first IBM Personal Computer, with no inherent networking capability and lacking even a hard disk drive, debuted in 1981. I don't recall exactly when we got the DG system implemented, but some history can be found here.
