Wildfire Statistics in the United States

This is part 1 of a three-part series on wildfires in the western U.S. Here's part 2, where I delve into more of the geographical statistics on unprotected lands. And here's part 3, where I examine the evidence regarding the impact AGW is having on wildfires in the American West.

The above graph above shows the increase in western U.S. wildfire acres burned since 1984 as well as the amount of increase that is estimated to be due to climate change. The paper it comes from[1] estimates that climate change has roughly doubled the acres burned in the western U.S. beyond what we would be experiencing without climate change. The reasoning here is complex, but what I want to do in this post is talk about the statistics themselves. The above graph comes from data published in a peer-reviewed paper[1] and shows up in the 4th National Climate Assessment,[9] but it has been heavily criticized for cherry picking.[7] We're told if we go back to the early 20th century, we'll get a much different picture of U.S. wildfires that will show the above graph to be alarmist and perhaps even deceitful. Here's a graph of the data as it was presented on the NIFC's website.[6]

The differences between the the last 40 years and the 1930s and 40s are pretty striking. They give the impression that even if the science is correct that climate change is doubling area burned, we're still much better off now than we were 80 years ago, and there's really nothing we should be concerned about.  In fact, recently the data before 1960 was removed from the NIFC website, and this prompted a conspiracy theory on the "Watts Up With That" blog that this data was removed because it contradicted the narrative that wildfires in the western US are becoming an increasing problem due to climate change.[2] Watts makes a big point about when the data were changed from wayback machine data, but he gets some of the details wrong. Let me try to clear it up. The following is a time line of updates to the site, but please keep in mind that the dates are when the website was crawled, not when the site was updated:

June 25, 2011  The data that was published was from 1960 to 2010 with the caveat that "Figures prior to 1983 may be revised as NICC verifies historical data. Stats after 1983 were compiled by states and agencies."

Jan 14, 2018  The data still only goes back to 1960 but the caveat is expanded. "The National Interagency Coordination Center at NIFC compiles annual wildland fire statistics for federal and state agencies. This information is provided through Situation Reports, which have been in use for several decades. 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 above prior to 1983 shouldn't be compared to later data."

March 31, 2018  The data that was published extended back to 1926. So the update with the 1926-1959 data happened between Jan and Mar 2018. There's this caveat: "The National Interagency Coordination Center at NIFC compiles annual wildland fire statistics for federal and state agencies. This information is provided through Situation Reports, which have been in use for several decades. 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." This data is located here until Jan 2021.

March 18, 2021 The data that was published is on a new site (the old page is down) and extends back only to 1983 with the caveat "Prior to 1983, the federal wildland fire agencies did not track official wildfire data using current reporting processes. As a result, there is no official data prior to 1983 posted on this site."

So generally speaking, the website showed 1960-present from 2011 to 2018 with a caveat. Then 1926-1959 was added from March 2018 to Jan 2021 with a caveat. From March 2021 to now the website only shows 1983 to present with an explanation. Watts' explanation for this is to jump to a conspiracy.

Clearly, wildfires were far worse in the past, and clearly, now the data tells an entirely different story when showing only data post-1983. The new story told by the sanitized data is in alignment with the irrational screeching of climate alarmists that “wildfires are driven by climate change”.

This wholesale erasure of important public data stinks, but in today’s narrative control culture that wants to rid us of anything that might be inconvenient or doesn’t fit the “woke” narrative, it isn’t surprising.[2]

Let's not jump the gun on this, though. Perhaps there's another, more rational explanation for this than that the NIFC wanted to hide data it previously had no problem publishing. After all, there were these caveats that were published about the data prior to 1983. Maybe there's something to learn about the data that will explain why the 1926-1959 data was added in 2018 and why the 1926-1982 data was removed in 2021.

Data Sources for US "Wildfire" Activity Data

To begin, we should note the different data sources used to compile this graph.[4] Above you can see that pretty much all the data up through about 1970 comes from USFS annual summary reports. Following 1970, USFS individual fire reports are added, and beginning in the 1980s multiple other sources of data were added to quantify the area burned each year.

Fires on Protected and Unprotected Lands

It turns out the source of the 1926-1959 data comes from the U.S. Forest Service, and the USFS published an evaluation of their own data from that time period. Above I posted a graph showing just how much of it comes from unprotected lands.[3] Notice how a huge proportion of the hectares burned before 1960 was on unprotected lands, while following 1960 it's pretty much negligible. Now let's look at where these fires on unprotected lands occurred.

Where Fires on Unprotected Lands Occurred

Notice that almost none of the hectares burned on unprotected lands come from the western US. Almost all of it comes from the southeast, and the vast majority of that comes from three states - Florida, Georgia, and Mississippi.[3] Notice it reports an average of ~10 million hectares per year burned over a 10-year period from 1927-1936 in Florida, Georgia, and Mississippi. That's a total ~100 million hectares or ~386,000 square miles. But the total area of those three states is only 173,000 square miles. So according to this data, the equivalent of the entire area of these states burned within their boundaries at least twice over that ten year period. This should cause any rational person to question what's actually being reported here. This is clearly not simply wildfire data in agreement with a definition of "wildfire" following 1960.

It turns out the definition of “wildfire” changes in the multiple sources used to compile the NIFC graph. Most of the areas burned in these unprotected lands were woods-burning - they were intentional, incendiary fires. They were what we now call “controlled” or “prescribed” burns, and they were intentionally set for controlling vegetation growth or clearing land for other uses. These fires are still done today, but they are not reported as “wildfires” unless a prescribed burn goes out of control. 

“In addition to the area-reporting bias in the USFS summary reports, analysts should be aware of other inconsistency and uncertainty in the wildfire activity estimates that are included, especially for unprotected areas. Intentional (‘controlled’) burning was used extensively for vegetation management on non federal lands, especially in the south-eastern US during the early 20th century. Although now used to a lesser extent (but on both federal and non-federal lands) in the US, intentional burning is not classified in the current reporting systems as ‘wildfire’ unless the controlled burn escapes and requires a suppression response. However, the early USFS wildfire activity summaries do include millions of hectares of intentional burning on ‘unprotected’ lands, which, until approximately the mid-20th century was viewed by the USFS as akin to wildfire, as something that should be prevented and ultimately eradicated (Pyne 1982).”[4]. 

“Millions of hectares of intentional burning in the south-eastern US are included as part of the annual estimates of wildfire activity on unprotected lands for several early decades of the USFS annual summary reports. It is inappropriate to compare, for example, the early total area burned estimates, which factor in ‘prescribed fire’, with the statistics in later USFS reports, which omit it, or to compare them with figures post 1998 in the NICC reports, which greatly underestimate levels of intentional burning in the US.”[4] 

The difference between datasets reporting true "wildfires" only and those reported in the USFS reports is significant. Short estimates that "areas burned intentionally in both agricultural areas and wildlands" may "account for 80% of the total area burned in the US in any given year."[4] Short concludes, "It is inappropriate to compare, for example, the early total area burned estimates, which factor in ‘prescribed fire’, with the statistics in later USFS reports, which omit it." The reported values in those earlier years are not reporting the same definition of "wildfire" used in recent decades. 

The NIFC agrees that the data before 1960 is not reliable. Here's what Randy Eardley at the NIFC told CarbonBrief.

I wouldn’t put any stock in those numbers. To try and compare any of the more modern data to that earlier data is not accurate or appropriate, because we didn’t have a good way to measure [earlier data]. Back then we didn’t have a reliable reporting system; for all I know those came from a variety of different sources that often double-counted figures. When you look at some of those years that add up to 60 or 70 million acres burned a lot of those acres have to be double counted two or three times. We didn’t have a system to estimate area burned until 1960, but it was really refined in 1983.[8]

Additionally, there are sources of bias in reporting that can't be consistently corrected in the early years. For instance, some burn-area estimates are based on fire perimeters, but that necessarily counts areas within the perimeter left unburned within the burn-area estimate. Satellite products show this estimation method causes an average overestimation of about 28% for area burned. This has prompted the NICC to adjust these estimates by a factor of 0.72 in recent decades. Likewise, some satellite products can overestimate areas burned because of the products' pixel size.[4]

So this is why the NIFC isn’t showing data prior to 1983. There are vast inconsistencies in the data reporting methods, even to what gets reported as a “wildfire,” and the accuracy of this data is questionable. There also a lack of consistent bias correction applied throughout the time series. It’s even likely that there was double and/or triple counting of acreage burned in the early USFS reports.[6] I don't know if it's possible for data analysts to untangle the data for 1926 to 1959, but given that most of the problems exist with data regarding unprotected lands, perhaps the best approximation of actual wildfire trends would be to limit the data to wildfires on protected lands. 

At the very least, we can say that the problem here isn't primarily why the 1926-1982 data was removed from the NIFC website. The real problem is, why was the 1926 to 1959 data ever added? You could argue that now that it's been published it looks suspicious that it's taken down. They could have just said something about that data to distinguish it from data following 1983. But of course that's precisely what they did when the data was on the website. They said, "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." That didn't stop people from using the data from the early 20th century and comparing it to data after 1983. So they took it down. I want an explanation for why it was ever added to begin with.

References:

[1] Abatzoglou and Wiliams, "Impact of anthropogenic climate change on wildfire across western US forests" PNAS October 18, 2016 113 (42) 11770-11775
https://www.pnas.org/content/113/42/11770

[2] Anthony Watts. "CAUGHT: ‘Inconvenient’ U.S. Wildfire Data Has Been ‘Disappeared’ by National Interagency Fire Center @NIFC_Fire"
https://wattsupwiththat.com/2021/05/13/caught-inconvenient-u-s-wildfire-data-has-been-disappeared-by-national-interagency-fire-center-nifc_fire/

[3] Short et al. USDA Forest Service Fire Data

https://www.fs.usda.gov/rm/pubs_journals/2020/rmrs_2020_short_k001.pdf

[4] Short Karen C. (2015) Sources and implications of bias and uncertainty in a century of US wildfire activity data. International Journal of Wildland Fire 24, 883-891.
https://doi.org/10.1071/WF14190

[5] Politifact. "U.S. acres burned each year are much fewer now — even in our worst years — than was the case in the early 20th century."
https://www.politifact.com/factchecks/2021/oct/15/heartland-institute/no-wildfires-werent-bigger-1920s-and-30s-today/

[6] "The Mysterious Wildfire Chart." ...and Then There's Physics. Oct 30, 2015
https://andthentheresphysics.wordpress.com/2015/10/30/the-mysterious-wildfire-chart/

[7] Bjorn Lomborg: https://twitter.com/bjornlomborg/status/1415338280233705478

[8] Zeke Hausfather. "Factcheck: How global warming has increased US wildfires."

https://www.carbonbrief.org/factcheck-how-global-warming-has-increased-us-wildfires

[9] NCA4. "Figure 25.4: Climate Change Has Increased Wildfire."

https://nca2018.globalchange.gov/chapter/25/#fig-25-4