California Wildfire Denial

Roy Spencer has joined the chorus of those denying the link between climate change and California’s wildfire crisis. But rather than share scientific insight, Spencer inflicts on us his brand of superficial, crude thinking through his inevitable confirmation bias, which, when examined closely, is an object lesson in how Roy Spencer got the label “climate denier.” The old-fashioned way: he earned it.

I took a look at relationships between California wildfire burned area, and the weather variables featured in the research on California wildfire from Williams et al.. Mine is a crude analysis compared to theirs, but yielded some interesting things.

We’ll start with total acres burned by wildfire in California since 1972 (the time span studied in Williams et al.).


Clearly there’s an upward trend, and in addition the fluctuations themselves — the wiggles up and down year to year — get bigger as the total area grows. We can “equalize” those fluctuations (a useful thing statistically) by taking the log-transform of the area burned, as did Williams et al.:

Clearly there’s an upward trend.

For climate variables I’ll use both high and low temperature in California averaged over the “warm season” as defined by Williams et al. (March through October), and the average precipitation during the “water year” (October through September). Precipitation shows no detectable trend during the time interval:

But temperatures, both highs and lows, have warmed strngly:

A crucial part of fire danger is how dry the potential fuel is. This depends heavily on something called vapor pressure deficit, or VPD. It’s the difference between how much water vapor the air can hold, and how much it does hold. When VPD is high, the air can suck moisture out of fuels with surprising speed. When VPD is high for sustained periods (as in, all summer long) it can dry out even the biggest fuels in heavily forested areas.

How much water vapor the air holds is its specific humidity. How much water vapor the air can hold depends only on its temperature. Higher temperature increases the vapor pressure deficit, that dries out fuels for fire, and that increases fire risk. Over the last several decades, temperature and vapor pressure deficit have both increased:

The increase in VPD has increased how quickly fuels dry out; it is, without doubt, one of the reasons California’s wildfire problem is so severe.

Of course VPD increase isn’t the only factor. If we plot the mean warm-season temperature against the precipitation during the water year, we clearly see that hotter years have more wildfire (red upward-pointing triangles for years with above-average area burned, blue downward-pointing triangles for below-average area burned, larger triangles for farther from average):

However, there doesn’t seem to be any additional correlation with precipitation. But if we make the same graph using the previous year’s precipitation, we see that the worst wildfire years tend to be those with both high temperature and lots of precipitation in the water year before the current one:

To test the idea quantitatively, I fit a model to burned area using only high temperature and precipitation. Using the current water year, the precipitation influence is nowhere near statistically significant. Using the prior year’s precipitation the significance is beyond doubt.

The “best” model I’ve found (in terms of AIC, the Akaike Information Criterion) is a model depending on time itself, a quartic function of both high and low temperatures, and precipitation during the preceding year but not the current year. The model compares to the data thus:

The model fit is impressive. I caution the reader that this analysis is no substitute for the more rigorous and far more thorough analysis of Williams et al. Nonetheless, the numbers make it clear that wildfire area burned is strongly related to temperature (high temperatures more so than lows) and the preceding year’s precipitation. It also depends on a time trend which does not represent any of the climate variables used. This additional trend could be due to many factors, including other climate changes (such as earlier snowmelt leading to further drying), but I suspect it’s more a reflection of some of the non-climate factors, particularly historical firefighting practices and increases in human-caused ignitions.

But the impact of climate change — specifically, of increased temperature via increased vapor pressure deficit — is evident. So too is the influence of the preceding-year’s precipitation, related to a significant amount of the variation but not the trend, because precipitation shows no trend (yet!).


If scientists have established any connection between climate change and wildfire, Roy Spencer means to dispute it. For example, he says this:


… the argument I’ve seen that excessive vegetation growth from a previous winter with abundant precipitation produces more fuel is opposite of the observation that fewer wildfires typically follow an unusually wet winter in California. They can’t have it both ways.

Spencer “supports” this statement by linking to an article — in the San Jose Mercury-News. That’s not a peer-reviewed scientific journal; it’s a newspaper. Apparently this is where Roy Spencer gets his “science.”

If he had bothered to read the science, he’d know that wildfire risk is greater when a wet year builds up fuels, then is followed by a dry year that dries them out. As Williams et al. say:


To evaluate observed and modeled trends in fire‐promoting interannual precipitation volatility (wet years that grow fuels followed by dry years that dry fuels out), we examined the running 10‐year frequency of “wet‐dry events,” which we define here (building off of Swain et al., 2018) as events in which a lowest 20% water‐year (October–September) precipitation total follows a highest 20% precipitation total in at least one of the two preceding water years.

Roy, you are just plain wrong, because you didn’t understand the science. You just spouted nonsense based on nothing but an article in a newspaper.

Then Roy Spencer graces us with this gem:


In 2018, a paper was published by a university research meteorologist and U.S. Forest Service (USFS) employees from three different USFS offices that describes a simple meteorological index related to wildfire risk. They call it the Hot-Dry-Windy (HDW) index, which is simply the product of (1) the surface wind speed times (2) the water vapor pressure deficit. The vapor pressure deficit uses the same information as relative humidity (temperature and dewpoint temperature), but it is a difference rather than a ratio, which better measures the potential of air to rapidly remove moisture from dead vegetation. For example a 10% relative humidity at 40 deg. F will have low drying potential, while 10% RH at 100 deg. F will have very high drying potential.

Vapor pressure deficit? Where have I heard that before? …

He goes on to emphasize the importance of wind in spreading wildfire, and implies that the windiness is the same as it ever was. He somehow fails to make the connection that higher temperature, an obvious and strong trend in climate change, is making VPD trend upward, and that’s one of the things making wildfire are burned trend upward. As Williams et al. showed.

As for trends in these variables, Spencer says this:


Unfortunately, the website does not provide any time series of the data over the last 30 years. But I can see the technique being applied to weather station data that goes back 50 years or more, for instance the formatted weather station data available here (which is where I got the Los Angeles airport data plotted above).

Until someone does this (if they haven’t already), I think it is a mistake to blame increased wildfire activity on “climate change”, when we don’t even know if there has been a change in the meteorological events most associated with major California wildfires: the intrusion of cool Canadian high pressure areas into the U.S. Southwest.

Williams et al. already did that. They blame increased wildfire activity partly on climate change. So do I.

Until Roy Spencer stops spouting nonsense without bothering to learn the science, until he frees himself from his blatant indulgence in confirmation bias, I think it is a mistake to trust him on any subject whatever.


Thanks to Geoff Beacon, Steve Lusky, and James Owens for very kind donations to the blog. If you’d like to help, please visit the donation link below.


This blog is made possible by readers like you; join others by donating at My Wee Dragon.


14 responses to “California Wildfire Denial

  1. Good job, thanks.

    I really like the red-high/blue-low painting of the observations.

  2. Thank you. This is a good antidote to the WUWT blog entry by Willis Eschenbach that was posted last year during the middle of the Paradise/Camp blaze, where he claimed that the California warming over the last 100+ years was only 0.02 C/decade (which was 10 times less than was reported in the scientific literature) and clearly much less than the high temperature trend shown in your post.

  3. See also Ficklin and Novick 2017, which looks at observed and future (modeled) trends in vapor pressure deficit in the US. California (and the US southwest in general) have had huge increases in VPD, especially in the summer and fall. See Figure 1 for maps.

    https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1002/2016JD025855

    • FWIW, I tried to post this link in a comment at Roy’s blog, but it hasn’t made it through moderation.

      [Response: Perhaps Roy is just too embarrassed by having his obvious nonsense exposed.]

  4. The last chart that you modeled is very impressive and clearly not a case of over-fitting (what can be overfit?). Did you happen to calculate an R2 or correlation coefficient for it? I am guessing a CC at least 0.8, which would be excellent considering how much fluctuation is in the data.

  5. Nice clear summary. Deniers gonna deny, sad. Meanwhile, this:
    ‘Ecological breakdown’: Greta Thunberg and youth activists rally as wildfires burn: The Swedish climate activist joined more than 1,000 people for an afternoon of youth-led climate action in Los Angeles
    https://www.theguardian.com/environment/2019/nov/01/great-thunberg-los-angeles-rally-paradise-wildfires and

    Fracking halted in England in major government U-turn: Victory for green groups follows damning scientific study and criticism from spending watchdog which, of course, is just a vote-getting pose by pro-fossil conservatives, but still, any port in a storm.
    https://www.theguardian.com/environment/2019/nov/02/fracking-banned-in-uk-as-government-makes-major-u-turn

  6. That is an impressive fit. One would expect the problem to be much more complicated, but maybe it is not or these other factors did not change much over this period.

    Precipitation shows no detectable trend during the time interval:

    The relevant trend would be in the year to year variability. By eye I do not see any, but it may be interesting to compute this.

  7. If I’m not mistaken, you are using the exact same dataset for exploratory analysis, model generation, and model validation. The model is not necessarily wrong but the AIC and “impressive” model fit wouldn’t have any meaning if this is the case.

    I could very well be wrong about your methods, I hope so!

  8. Hi. I recently started a facebook group. The group is called All the Children of the World. My plan is to make it a place where children from anywhere in the world can see the different walks of life their peers have. If you are interested in this group please contact me or look it up on facebook. Thanks, Ray

  9. Year to year variability of precipitation in California is related to water vapor and warm water export from the tropical Pacific. This goes through wild swings because of ENSO. ENSO response to warming climate is still uncertain. Very impressive relationship between previous years precipitation and acres burned.

  10. A Nun said:
    “If I’m not mistaken, you are using the exact same dataset for exploratory analysis, model generation, and model validation. The model is not necessarily wrong but the AIC and “impressive” model fit wouldn’t have any meaning if this is the case.”

    The impressive point about the fit is that there is very little in terms of wiggle room for making adjustments to the fit. Those are the degrees of freedom (DOF) and there are essentially very few — perhaps the use of the previous year’s rainfall is one, but that it based on a real physical process.

  11. In case you’ve not seen it, Stefan Rahmstorff drew attention to this paper relating the increase in precipitation required to offset the fuel dehydration caused by a temperature increase of a degree. https://link.springer.com/article/10.1007/s10584-015-1521-0

    • From the abstract:

      When temperature increases we find that for every degree of warming, precipitation has to increase by more than 15 % for FFMC (Fine Fuel Moisture Code), about 10 % for DMC (Duff Moisture Code) and about 5 % for DC (Drought Code) to compensate for the drying caused by warmer temperatures. Also, we find in terms of the number of days equal to or above an FFMC of 91, a critical value for fire spread, that no increase in precipitation amount alone could compensate for a temperature increase of 1 °C.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s