Guest Post by Willis Eschenbach
Inspired by the work done by Robert Rohde attempting to link May to October temperatures and rainfall to fires, I thought I’d take a look at the acres burned over the years. Rohde compared the rainfall and temperature records and highlighted the largest fires. However, this gives only a few data points. I wanted a larger view of the situation.
So instead of major fires, I looked at the areas burned every year, which are available here. There is complete data from 1959 to 2016, and the last two years are available here and here.
The first thing I did was run a multiple regression on the data, using both May to October temperature and May to October rainfall to see how well they would predict the area burned. To my great surprise, I found out that rainfall is not significantly correlated with the area burned. Here is that result:
Estimate Std. Error t value Pr(>|t|) (Intercept) -13393785 2502402 -5.352 1.61e-06 Temperature 203834 35791 5.695 4.52e-07 Rainfall -46812 35591 -1.315 0.194
Temperature is significant (right-hand column, p-value 4.52e-7), but rainfall is far from significant (p-value = .19). So I ignored rainfall for the rest of the analysis.
Next, I graphed the acres burned, and ran a linear regression on the data. Figure 1 shows that result:
Figure 1. Total areas burned by year, 1959-2018 (red line) and linear least squares trend line (blue line).
Note that the p-value of the line is quite good (right column, p-value = .00000004). The R^2 value (bottom line) shows that the straight line explains 41% of the variance in the acres burned.
Then I looked at the connection between temperature and acres burned. Figure 2 shows that result:
Figure 2. Total areas burned by year, 1959-2018 (red line) and acreage estimated from the variation in May – Oct temperatures (blue line).
Curiously, that looks a lot better than the straight line … but note that there is only a slight increase in the amount of variance explained (44% variance explained by temperature versus 41% for the straight line). This proves once again that our eyes are tuned to see patterns even when none are there … consider the constellations of the night sky as a prime example.
Finally, I looked at the errors in the temperature based estimate of the acres burned. Figure 3 shows the difference between the temperature-based estimate of the area burned and the actual acreage burned.
Figure 3. Errors of the estimate.Total areas burned by year, 1959-2018, minus the acreage estimated from the variation in May-Oct temperatures. The red line is a seven years Full-Width Half Maximum (FWHM) Gaussian average of the data. The vertical dotted blue line shows that in 1994, logging was stopped in much of the state in a vain attempt to protect the Spotted Owl.
Here is the oddity of the Figure 3 graph. In the first part of the record, up to the early part of the 21st Century, the temperature generally overestimates the acres burned.
But since then, the temperature has greatly underestimated the number of acres burned.
This is clear evidence that the recent large wildfires are not due to the variations in temperature as is widely claimed.
Conclusions
• Using variations in May to October rainfall do not improve the estimate of the acreage burned. In other words, May to October rainfall doesn’t add anything to an estimate done using May to October temperature alone.
• The variations in May to October temperature are only slightly better than a straight line in estimating the variations in area burned.
• The recent very large areas burned are not the result of variations in May to October temperature. As I pointed out in my last post, the decade over decade changes in temperature are nowhere near large enough to explain the recent increase in the area burned. We must look elsewhere for the causes of these large fires.
h/t to Steven Mosher for pointing out the Rohde analysis.
[UPDATE] Some folks didn’t like the data I used, so here’s some more:
I’m home now, and the smoke is not too bad. Not as bad as it was in the Central Valley or in San Francisco on the way back here. We’re in the yellow area on the California coast north of San Francisco. Smoke map available here, click on “Vertically Integrated Smoke” or “Near Surface Smoke”.
Figure 4. Smoke map. Red is the thickest smoke. The Camp Fire is burning north-east of Sacramento.
My best wishes and hopes for the future for all of those affected by the fires, and my condolences for those who have lost friends, family, or homes.
Regards,
w.
PS—My usual request. When you comment, please quote the exact words you are referring to, so we can all be clear on what you are discussing.
Perhaps the increase of forest fires in settled ares of CA, as well as Greece and other areas with a Mediterranean climate, has to do with the eucalyptus tree and its continually growing popularity. This tree is extremely flammable, as some in Australia are aware of. This tree has all sorts of advantages, including rapid growth, repelling insects and even having a pleasant aroma, so we might as well just live with its flammability.
Eucalyptus is not native to California. Some was planted in the past by farmers and is occasionally used in landscaping for homes. There probably isn’t any at the location where the fire started. Most of the trees were native pines and fire trees which are also extremely flammable.
A very serious situation for many. Please be safe. I ran into a series of videos posted by Juan Browne on his Blancolirio youtube channel. He has some interesting observations including a video of the area where the fire started two days before. Apparently the area had been burned and most of the forest growth was destroyed leaving basically dry scrub growth in the canyon. He has some interesting observations and is at risk himself. He gives a very reasoned account unlike most of the media. Here is the first of his reports. Others can be found on his channel. By the way he has probably the best series of reports/videos on the Oraville Dam disaster and its reconstruction. Well done and very interesting.
Yes Juan’s coverage of Oroville has been great as has his coverage of the recent fire (he lives about 30 miles from it). As an ex-firefighting pilot (currently an airline pilot) he brings a great insight into the Ca fires, he’s posted on this subject over the last few years. He was actually in the mountains near where the fire started few days before and showed how dry the ground was and commented on the extended dry season this year (only 0.25″ since the end of september, last year by this time they’d had 5″).
Maybe it would correlate with a number of regulations? I need a permit to burn my dead wood, and I can do it only during approved “burn days”. I’ll speculate that there may be 25 burn days in a year.
“Maybe it would correlate with a number of regulations?”
You would probably get a better correlation with the budget for controlled burns and thinning. Funding for the US forest service is very low for this. California has been increasing the budget for this but fire fighting costs has increased even faster than the budget. Most of the money for fire fighting this year has already been spent in after only a couple of months. It currently looks like California will need to spend about 1.5 billion a year on fire fighting, controlled burns and thinning.
Willis,
You said, “…the straight line explains 41% of the variance in the acres burned.” Not to be pedantic, but that should be 40.1%.
The fall rains are a few weeks late showing up this year, and combined with excessive fuel loading from years past, have now combined with an ignition source to create a potential monstrous fire. It could have happened anywhere else, but unfortunately it conspired within an area of rural and city interface. I am so sorry for the people suffering through this. This is very tragic and is my worst nightmare, living within my own managed woodlot forest. When it is hot and dry, even after a few dry weeks in May/June, anything is possible with the right wind and ignition source.
If there had been a decent fall rain in October or early November, these present conflagrations would have probably waited for another opportunity to burn. Sooner or later, it is the nature of everything to burn in a firescape evolved ecology. It is tragic that we are now in the mix with human habitation. We either have to adapt with human intervention in reducing fuel loading/ignition sources or suffer the consequences. Much of the land is not commercial forestry so it all can’t be blamed at the foot of logging or the spotted owl, although that could certainly be a huge factor where forest fires actually happen where commercial forestry could be happening and at least offer some management and reducing potential areas of burn.
Perhaps we should be looking at artificial cloud seeding to coax out any possible rains, when there is any opportunity to do so? I know in the Philippines, where the Province of Bohol in the Central Visayas has been in prolonged drought and has begun targeting cloud seeding to fill specific dam reservoirs that are utilized for irrigation of their rice fields that are failing due to prolonged drought. Over several years of similar periodic droughts, they have managed to squeeze out some rains to localized areas impounding water behind dams for irrigation releases, enough to justify to keep trying with this technique. They are extremely poor there, so I don’t think they are wilfully wasting money up a rope doing something for nothing. I realize you need a specific cloud set and humidity to come along that is conducive to artificial cloud seeding, but when it does, there is a hope of getting some much needed precipitation. Utilizing 15-30 micron vacuum packed salts as cloud seed nuclei seeded by a small dual prop airplane. Just a thought…if it worked a few times, might be worth exploring. Anything at this point should be worth taking a closer look at. A few interesting local stories from the cloud seeding efforts in Bohol, Philippines below.
http://www.boholchronicle.com.ph/2018/11/14/cloud-seeding-in-bohol-suspended-amid-rains-due-to-lpa/
http://www.boholchronicle.com.ph/2018/09/16/cloud-seeding-awaited-to-water-bohol-farmlands/
http://www.boholchronicle.com.ph/2018/09/03/p2-3m-allotted-for-cloud-seeding-to-save-rice-harvest-in-bohol/
There has been at least one year recently when early spring rain led to a lot of grass in an otherwise dry summer. I think it was February that rained so hard there was flooding in California, followed by fires.
Yes there was exceptional rain in Feb 2017 which led to the damage to the Oroville spillway (over 12″), most of the dams in that region of Ca filled up and of course lots of grass growth during the following spring. Leads to lots of potential fuel.
My research indicates that you will find a high degree of correlation between fuel load and total areas burned. U.S. total timber volumes have increased by over 60% since 1953.
Paul, the best source for this kind of data is the FAOSTAT Forestry Database.
Regards,
w.
My sympathies for Anthony and the community that surrounds him.
NWS shows ytd rainfall totals for Chico 6”, Redding to the north, Sacremento to the south both at 13”. The station data for Paradise being somewhat east at Yarbo Gap at 39” (?)
Thousand Oaks is 7”, Malibu is 10”.
My recollection is since the last rain event in early April ‘18, SF Bay Area has had but one day of trace rain in the subsequent 7 months.
CA ‘16-17 water year anomaly wet, massive fires. ‘17-18 water year anomaly dry, again massive fires.
“NWS shows ytd rainfall totals for Chico 6”, Redding to the north, Sacremento to the south both at 13”. The station data for Paradise being somewhat east at Yarbo Gap at 39” (?)”
Redding, chico, Sacramento are all at low elevations and in the rain shadow of the coastal mountains. All are in semi dessert environments. Rainfall varies a lot year to year normally in these dry environments.
Yarbo gag is further east in the low Sierra Nevada foot hills. 39″ is probably close to normal for that location. Paradise is further east and higher up so it should normally get a little more rain than Yarbo gap. This amount of rain variability is normal in California.
The wettest locations in the state are coastal and sierra Nevada mountains. in the 2017 /2018 winter snow fall in the mountains was a less than normal but the rainfall was a little above normal. End result was that 2017/2018 winter was about average to little below average. But since everyone focuses on the snow report and weather forcasters, and reporters are calling last winter a drought. It was not.
The current california drought monitor website lists the state at D0 abnormally dry, the lowest drought conditions.
https://www.drought.gov/drought/states/california.
If you look at hte conditions 3 and 4 years ago when we really did have a drought the difference between then and now is dramatic.
Many people don’t simply understand California weather. Unlike the east cost which gets rain fall almost every month of the year with high humidity levels in the summer. California typically gets zero rainfall from jun through September with almost no humidity. May and October are transition months and frequently get no rain but occasional they do. November is typically when the rains really start but it is still not unusual for the rain to start at the end of the month. Currently it looks like we might get rain around thanksgiving. So 7 months without rain is about normal for the state. Unfortunately October and November are typically the months when high winds are common.
Sorry, I meant Jarbo Gap and ytd precip 35 not 39 inches. Paradise at 1800’. JBGC1 Elevation is listed at 2500 ft, not snow pack level, and the lat long google map for it is east of both Paradise and rte 70 between Pulga and Parkhill. Well within boundaries of the fire.
My point, poorly made, was the fire once ignited burned through areas relatively “drier” at the western Chico end and also 3x “wetter” areas north and east. Longer term drought or no, not my issue, the relevant water year anomaly ending Oct 2018 was between 20% and 25% below normal.
https://www.climate.gov/news-features/event-tracker/dry-2018-water-year-comes-end-california
I completely agree that until one lives here it is difficult for an easterner’s mind to fathom the monotony and creeping dessication of our dry season.
Willis: interesting analysis. But here’s a thought… Long periods of drought will cause some trees severe stress. There may be several years of lag before those trees completely die, fall, dry, and provide fuel for a fire.
This leads to the unremarkable conclusion that removing dead trees and vegetation from fire prone areas would be prudent.
For the record, I don’t think temperature or rainfall is driven by CO2!
May to October rain seems limiting. Note that I’m coming from the perspective of a Pacific Crest Trail hiker where winter snow pack combined with its subsequent melt will undoubtedly influence the dryness of the forests, at least in areas subject to snow accumulation or melt runoff. An example was 2017 – a very heavy snow year in the Sierra, with a late melt off (which actually caused fatalities as a couple hikers died trying to cross melt swollen creeks / rivers). High snow plus later melt in a given year would tend, I would suspect, to delay the drying of the forest into fire prone conditions.
Additionally, i suspect there is some lagging effects between precipitation (over an entire water year, October to October) and subsequent fire as the accelerated vegetation growth from a wet year, or series of wet years, doesn’t necessarily burn until following dry year(s). I suspect this effect also applies in areas where the predominate vegetation grows on the annual time scale – a wet winter / spring followed by an abnormally dry summer and fall would result in bad fire conditions that year (sage brush, grasses, manzinita, etc). This would be in contrast to mature forest where I would suspect the pattern would be on longer time scales – at least the following year.
As pointed out already it doesn’t rain much in California in the summer and fall so it’s not surprising there’s little or no correlation. I doubt you’ll find a statistical correlation even using winter rainfall totals although it’s obvious that wets years increase the amount of fuel and drought creates higher risk of major wildfires.
From what I’ve read fire was a common occurrence 100+ years ago with areas burning every 15 to 25 years on average. Ancient giant sequoia show this pattern for example. That limited catastrophic fires by reducing fuel load and limiting fuel ladders that enable fire to reach the crown and kill mature trees.
It seems to me that if we knew the average amount of acreage burned annually 100 years ago and compare it to the annual average over the last 25 to 50 years then you could calculate the annual unburned acreage that is building up in the state.
The bottom line is that fires aren’t unusual in California. They’re a part of the natural ecosystem. The issue now is how we protect people and property and manage the urban wildland interface along with managing build up of fuel loads.
Willis :
Where is the graph of underbrush in State and National forests?
As I have said on this site before: When the pioneers went to California they had to carry wood for fires as the followed the Platte River across Nebraska as there was no fire wood within miles of the river. Today there is almost a forest along the entire length of the Platte river. The present areas of forest can be confirmed by any internet satellite map service.
As you zoom in you will note that most green areas in NE, UT, WY, KS, etc. are around center pivot irrigation systems. Where are the studies on their impact on rainfall, Temperature, Flora, Fauna, etc.
Maybe a comparison of how much acres of prescribed burns, forest thinning, and money spent on forest management would be in order. My guess that they do not do enough and that exacerbates the problem.
Forest management like that is expensive. It requires access, which can be difficult in mountainous regions. Thinning is not economically viable in some forests, especially if it means removing the material. Prescribed burns require appropriate conditions. Who’s going to pay for it, when state and federal forestry programs are often underfunded as it is?
Pay me now or pay me later.
Trump’s response to the fire is to threaten to stop all federal funding:
“There is no reason for these massive, deadly and costly forest fires in California except that forest management is so poor. Billions of dollars are given each year, with so many lives lost, all because of gross mismanagement of the forests. Remedy now, or no more Fed payments!”
No reason. Drought has nothing to do with it. And temperature increases associated with climate change have nothing to do with drought. That’s brilliant.
But hey, adaptation is better than prevention! So who cares about a few fires? People will adapt.
Kristi, you are being deliberately dense. The threat was contingent, that unless California changed their wildlands policy, the Feds would no longer pay for the consequences of their folly.
RE Kristi:
**No reason. Drought has nothing to do with it. And temperature increases associated with climate change have nothing to do with drought. That’s brilliant.**
What would be brilliant if Kristi could show that temperatures have increased aka “climate change” aka global warming.
Also it would be brilliant if anyone could fine a scientific study that measures warming caused by CO2.
Tom Halla,
It’s still a threat. Maybe a more appropriate reaction would have been to say, “We are going to review the federal funding for forestry lands in California, and see whether more funds can be shifted to measures which will increase fire prevention and control to keep populated areas safer.” 57 % of forests are owned and managed by the federal gov’t. 40% are owned by private individuals and companies, and native Americans.
Trump assumes that the cause is forest mismanagement. While that may play a role, it’s not the only cause. Besides, CA is doing something about it already.
“Westrup, the governor’s spokesman, pointed to the passage this year of a bill intended to reduce wildland fire danger.
“Under the bill, SB901, the state will dedicate $1 billion over five years to fire-reducing efforts such as clearing brush and setting prescribed burns. It also eases restrictions for clearing drought-killed trees from the state’s forests.
“Brian Price, president of the 30,000-member California Professional Firefighters, called Trump’s first tweet “ill-informed, ill-timed and demeaning to those who are suffering, as well as the men and women on the front lines.”
“At a time when our every effort should be focused on vanquishing the destructive fires and helping the victims, the president has chosen instead to issue an uninformed political threat aimed squarely at the innocent victims of these cataclysmic fires,” Price said in a statement.
“The president’s assertion that California’s forest management policies are to blame for catastrophic wildfire is dangerously wrong,” he added. “Wildfires are sparked and spread not only in forested areas but in populated areas and open fields fueled by parched vegetation, high winds, low humidity and geography.”
Who’s the one that’s dense?
If one actually pays attention to what California has done on wildlands management, it mostly rejects controlled burns and logging. Blather about environmental preservation presumes an original state that never existed, as the area has been inhabited since the end of the last ice age, and the Indians definitely practiced active forest and wildlands management.
The environmentalists have the notion that management is “preservation” of something that never existed.
On a bit of consideration, I may have been obscure. The fantasy of the environmentalists is that the Americas ever had a stable ecosystem that did not involve active measures by people. As people have been in the Americas since before the ecosystem settled down after the last ice age, there has always been human influence, and purposeful at that. Notably, there have been archaeology studies done in California, that the local tribes maintained as open woodland what became overgrown scrub after becoming a state park, with the interval being a cattle ranch, which had much the same effects as routine fires.
Gerald Machnee,
4https://www.epa.gov/climate-indicators/southwest
Tom Halla,
I’m not an environmentalist; it’s always irritated me when people suppose I am one. I agree with you about the influence humans have had on forests. The fire suppression humans have done during the last century has been a huge mistake, in most cases – but in others, it’s necessary; this is a drawback of the formation of permanent settlements in forested areas.
I’m not against logging, but I’m in favor of doing it wisely. There is so much I could say about this, it would probably bore you. Logging is a very important industry in MN, where I live, and I worked on developing a manual for foresters for the DNR.
Forest management for fire prevention is a different thing altogether. It’s costly. Controlled burns can be dangerous, especially in the middle of a drought. Thinning is sometimes counter-productive.
Besides, it’s not just forests that burn.
There is no easy solution, which is why Trump’s remarks were so widely criticized by those who know about wildfire. A better way of looking at the problem is from the side of development and construction. The urban-nature interface has to be more closely examined, and communities have to start seeing wildfire as a potential rather than relying on vegetation management to protect them.
Besides, forest thinning can actually lead to higher fire risk, since the understory gets more light (which can affect fuel loads as well as drying) and increases wind, which adds to drying.
“But since then, the temperature has greatly underestimated the number of acres burned.
This is clear evidence that the recent large wildfires are not due to the variations in temperature as is widely claimed.”
No, you have given no such evidence. You have only shown that there is not a perfect correlation between temperature and acreage burned. Temperature could still be one of the factors – and if it explains 44% of the variance, that’s no unlikely.
You also didn’t look for a lag time in precipitation’s effects, as others have noted.
Those who look for one or two factors in increased wildfire are not allowing for the fact that fires can have many influences. This is as much an error of those who blame all increase in fire on climate change as those who say climate change isn’t a factor. Willis, your analysis suggests to me that temperature is likely one factor, and it puzzles me that you deny it.
Kristi, I’m gonna pass. You always come in here, deny any and everything on often specious grounds, whine about how I haven’t done something that you could easily do yourself and report back the results … and then you never listen to a dang word that I or anyone else says to you in response.
As a result, discussing this stuff with you is boring as hell, and I’ve learned my lesson. Go bother someone else, please, I’m not interested in the slightest in your opinions.
w.
I didn’t ask a question, anyway, Willis, I made a comment. Your statistics have interesting results, and it’s weird that you don’t see it. Fine if you have nothing to say in return.
I have just as much right to comment here as anyone else.
Pass … not interested.
w.
“your analysis suggests to me that temperature is likely one factor, and it puzzles me that you deny it.”
Realistically, temperature doesn’t have that much to do with it. The key factors are fuel load, oxygen supply and dryness of fuel. I’ve burned large outdoor fires when it was single digits and when it was 90F, and they burn the same. Fire affects the local temperature, more than vice versa. There’s actually more oxygen in cooler air. If you want a really hot open fire you add oxygen, not heat.
icisil,
Temperature has an effect on drought and moisture in the vegetation.
Willis –
Temperature and Acres Burnt are both rising, so I believe there will necessarily be a correlation between them, even if they are unrelated phenomena. I think you need to de-trend both series first before doing the correlation analysis.
I did try that, Russell. It doesn’t change anything, except of course the correlations get worse.
w.
Reassuring, thanks for the response! R.
My thought is that there may be a correlation with the combination of spring rain and length of summer/fall drought.
As others have pointed out, May-October rainfall in CA is insignificant to the calculations, as it insignificant in itself. The wet season is in the wintertime, and if rains are abundant, forest fire numbers will be vastly reduced for the following year, as the moisture resides in the soil and vegetation. Where has every one been for the last `seven years? California has been in a severe drought (except for 2015/6-17), making the forests a tinderbox. During winter 2015-17 Northern-Mid CA had heavy rainfall but not the Southern part which was still stuck in drought conditions–just as they are now, and the South had massive fires in 2016-currently. No matter how important sensible wood clearing is, that is not the main problem.
In parts of Europe, there are no forest fires because the weather system is very dynamic with regular wet/dry cycles (daily, monthly, yearly); the atmosphere is very healthy–blue, lack of “smog”, and does not suffer from dry seasons nor drought. One cannot even start a forest fire if one tried as the trees are full of sap and water. One can burn yard trimmings during winter season without prohibition, and I have seen one of these fires “over jump” its area and begin to smolder the surrounding bushes. It sizzled for a while and then went out because the bushes were too moist inside.
In summary, it is not land management, nor hypothetical global-warming, which feeds these fires, but the lack of a healthy weather system which naturally provides moisture by dew, humidity, and rain.
al,
It’s not global warming, but lack of a “healthy weather system”? Are you aware that dry areas getting drier is exactly what is predicted by global warming? And that temperatures are part of drought?
It’s not “hypothetical” global warming, no. It’s actually global warming, regionally expressed in CA, as Willis’s data show.
It’s striking that people will use such convoluted reasoning to deny that climate change has anything to do with wildfire. It’s not the whole story, but it can’t be eliminated as a factor.
“Dry areas getting dryer have been predicted by GW?” Of course, doomsday scenarios have been the mainstay of Global cooling/warming/change since its inception. Just one favor Kristi, please step outside in Houston with your Fall clothes and wake up to reality!
PG&E loses 25% of market value today alone with concerns on causing #campfire Paradise fire. PCG closing price today 25.59 USD, down −$7.13 today on a volume of 52 million shares. Stock price had a year to date high of about $55.63 a year ago and as recent as recent as Nov 7/18, down from about $48.80. Bonds are crashing as well, with the possibility of a bankruptcy. Facing $17.3 Billion in claims for 2017 fire season. Assuming even higher for 2018, especially if #campfire is deemed to have been started by PG&E.
Seeking state intervention and aid for billions in a possible bail out scenario if claims exceed insurance payouts. Market seems to be pricing in a guilty verdict already. Big, big troubles for PG&E. And should be for the Governor’s office as well, but I doubt anything will happen on that front.
https://finance.yahoo.com/news/pg-e-plunges-draining-credit-144537343.html
https://www.cnbc.com/2018/11/14/pge-plunges-20percent-after-disclosing-an-electric-incident-just-before-fire.html
https://www.marketwatch.com/story/pge-stocks-and-bonds-hammered-on-worry-it-may-be-liable-for-california-wildfire-2018-11-14
My thoughts go out to all in Ca affected by this tragedy.
The devastating bushfires in Victoria, Australia, in 2009 burned over 1 million acres in a few days, based on about seven major fires throughout the state (more than 300 individual fires). The weather was a contributing factor on with the main day of destruction with 47 degC temperatures (115 degF) and 90km/hr winds (56 miles/hr) winds. 180 people lost their life and many more were injured. Whole towns were wiped out.
The ignition causes of the fires were determined to be a variety of things : power line arcing, stupid use of machinery, some lightning and would you believe it, arson.
But the underlying cause of the severity of the fires is believed to be years of failure to clear forest fuel loads through controlled “cool burns” in winter, especially on government controlled land, and inappropriate housing developments in heavily wooded areas. Recent changes in several local council regulations stopping general fire safety practices such as clearing trees or bush on private land (mainly due to green ideology) was also a factor, and many regulations were changed back after the tragedy. The importance of fire breaks in forests and cleared zones around houses was also identified to be a significant factor in fire control and limiting damage.
The 2009 tragedy was a wake up call for Victorians that resulted in changed regulation and fire safety preparation. But Victoria has major fires every few decades, (1939, Ash Wednesday 1983, 2009) so we just wait for the next one.
Which is less expensive:
1) Laying out miles and miles of irrigation tubing and sprinkler systems throughout the forests to keep them moist- or some kind of plan, even flying over and dropping water before the fires start
2) Annual fires that destroy complete towns and thousands of acres of trees?
Joe G.
1. California does not have the water available to irrigate the forests. They barely have enough water for humans.
2. Trees grow back after fires. Towns need to protect themselves with cleared areas surrounding them.
“Rohde compared the rainfall and temperature records and highlighted the largest fires. However, this gives only a few data points. I wanted a larger view of the situation.”
A “few data points”? Since 1895, 9 of the top ten largest, and 9 of the top ten most destructive fires have been since 2000.
There are way more than a few data points here. There are two really interesting graphics – people should take a look if they haven’t already.
Looking for the data source for acres burned
usually willis includes his data.
I did find this WARNING about using the data
‘Due to the changes in data
collection, methods, and systems over the years, information may not always
be comparable and data may be of differing accuracy or completeness.”
Looking at these reports it is unclear what figures Willis used.
I will note after reading some of these reports that they often warn
due to the hot and dry weather…… we got a lot of burning
My suspicion is that linear regression is not the best way to go at this.
I have some thoughts, if Willis has the acreage burned data for show an tell
Hey, Mosh, caught your question on Twitter but I’d missed it here.
The area data is here as a CSV file.
w.
Hmm
http://www.fire.ca.gov/downloads/redbooks/1959_BW.pdf
Also
‘CAL FIRE’s firefighters, fire engines and aircraft respond to an average of 5,400 wildland fires and answer nearly 192,000 calls for other emergencies each year. During 2007 CAL FIRE responded to 3,610 wildfires within its Direct Protection Area. This represents a decrease of 33% below last year and 33% below the 5-year average. These fires burned 434,667 acres which is 49% above last year and 54% above the 5-year average. Up until October, California fire season had been relatively light. Seasonable conditions were seen throughout the state through the middle of October.
That all changed as forecasters predicted a strong off shore flow beginning October 21, 2007. CAL FIRE and other fire departments began to preposition staff and fire equipment throughout Southern California. On October 21st, numerous wildfires sparked across Southern California. With the combination of strong winds, low humidity and dry conditions, the situation was set for a disaster. Fueled by dry vegetation and strong Santa Ana Winds, firefighters battled several raging infernos. Nearly one million residents were evacuated. This was the largest mass evacuation in California history. Ten people lost their lives and over 510,000 acres were charred from the fire siege. The last of the raging wildfires were not fully contained until early November, 2007.”
yes, it seems these fire fighters know nothing about the relationship between hot, dry, offshore flow
and fire.
Bottom line. If you regression is not showing a relationship between hot and dry and worse fires
then
A) you regression or data is missing something OR
B) fire fighters dont know jack
Steve, all I can tell you is what the data says. I suspect that the issue is that fires don’t respond to climate … they respond to WEATHER. And not to weather that lasts a month. If the Santa Ana winds blow for a few days, it may not change the monthly average temperature, but certainly may lead to many fires.
And since California gets very little summer rain, I’m not sure that May-October total rain will be relevant, and the data says it is not. As with the temperature, a week of hot, dry Santa Ana winds is very different from a week of normal winds blowing off of the ocean … and although they affect fires very differently, neither of them contain summer rain.
As to your comment that “these fire fighters know nothing about the relationship between hot, dry, offshore flow and fire”, that’s a huge strawman that you are trying to set on fire. Point out anyone who has said that hot dry Santa Ana winds don’t increase the fire risk.
My point is simple. We have no evidence that “climate change”, whatever that might mean in Governor Brown’s brain, is leading to more forest fires. And we have no evidence for Robert Rohde’s claim that May to October temperature and rainfall controls fire risk.
w.
I would not say we have no evidence.
in fact the evidence is in your residuals
First things. Linear regression is probably the wrong tool. One reason why is the variables
may have thresholds. For example, if 2.45 inches of rain or more is enough to depress a fire season somewhat you may not see a linear response difference between say 3 and 4 inches or between 4 and 5 inches.. wet may be a threshhold
Same with temperature. there may be threshholds
So rather than just assume a linear model ( although doing one first should show you what
you miss) you can do a nonparametric recurvsive partitioning model.
basically tree regression. There are many options in the machine learning space, I like ctree
random forest is another option, meh.
Anyway, go ahead an plot your lm(Acres ~temperature) and plot it. Pretty sure you did
Pretty sure you saw what I see. those few points that stick out over 1m acres burned? are they dry?
Any way, in the tree regression this becomes pretty clear
A simple partioning regression against temperature shows 3 distinct temperature categories
=68.4 & =69.1
Essentially the tree regression will tell you to use these categories of temperatures
as your predictor…
Your R^2 will increase and your model will be a better predictor. You can think of it this way
Any value between 68.4 and 69.1 is going to predict the same acreage burned.
Moving to 4 categories you can get a little bit better prediction, (r^2 .53 ) but what I see from three categories and the residuals from the simple linear regression tells the story.
When you do a tree regression you will see a few outliers in every prediction interval.
confirming the story
The outliers are typically drier years.
In short, the problem is you have evidence of the dry years being more problematic
in every temperature category, but there are very few data points, so in a linear regression
this information just looks like noise/outliers. Of course, you can just call them outliers
and say they mean nothing, or you could look at the outliers and ask what they teach you
in the 70 or so years you have there basically isnt enough data of dry years to influence the simple regression.. put another way.. the hot dry years have more fires than your model predicts.
Same data, different way of interpreting it.
Outliers sometime have signal
I would not say we have no evidence.
in fact the evidence is in your residuals
First things. Linear regression is probably the wrong tool. One reason why is the variables
may have thresholds. For example, if 2.45 inches of rain or more is enough to depress a fire season somewhat you may not see a linear response difference between say 3 and 4 inches or between 4 and 5 inches.. wet may be a threshhold
Same with temperature. there may be threshholds
So rather than just assume a linear model ( although doing one first should show you what
you miss) you can do a nonparametric recurvsive partitioning model.
basically tree regression. There are many options in the machine learning space, I like ctree
random forest is another option, meh.
Anyway, go ahead an plot your lm(Acres ~temperature) and plot it. Pretty sure you did
Pretty sure you saw what I see. those few points that stick out over 1m acres burned? are they dry?
Any way, in the tree regression this becomes pretty clear
A simple partioning regression against temperature shows 3 distinct temperature categories
=68.4 & =69.1
Essentially the tree regression will tell you to use these categories of temperatures
as your predictor…
Your R^2 will increase and your model will be a better predictor. You can think of it this way
Any value between 68.4 and 69.1 is going to predict the same acreage burned.
Moving to 4 categories you can get a little bit better prediction, (r^2 .53 ) but what I see from three categories and the residuals from the simple linear regression tells the story.
When you do a tree regression you will see a few outliers in every prediction interval.
confirming the story
The outliers are typically drier years.
In short, the problem is you have evidence of the dry years being more problematic
in every temperature category, but there are very few data points, so in a linear regression
this information just looks like noise/outliers. Of course, you can just call them outliers
and say they mean nothing, or you could look at the outliers and ask what they teach you
in the 70 or so years you have there basically isnt enough data of dry years to influence the simple regression.. put another way.. the hot dry years have more fires than your model predicts.
Same data, different way of interpreting it.
Outliers sometime have signal
ah shit temperature ranges got screwed up
Thanks, Mosh. If you think that tree regression is the right tool … why are you recommending that I use it? If I do that, knowing little about tree regression using machine learning, I’m sure to get roasted.
One problem is that we don’t have much data, just 1959 – 2018 for the fires. Once you divide that up into say three threshold groups, we’re down to ~ 20 data points in each group. With three groups, the Bonferroni correction says to get a p-value of 0.05 we need to find an actual p-value of 0.05/3 = 0.017. We also need to correct for using just half the rainfall data, so now we’re down to needing to find a p-value of 0.012 …
In any case, here’s all of my data as a zip file entitled “Fire Data.zip”. Please let us know what you find.
Best regards,
w.
Steven Mosher,
Am I missing something? Does Willis’s analysis not show that 44% of the variance of acres burned is explained by temperature, and a very low p-value? To me that seems like a pretty good signal, considering all the other factors that come into play.
Rohde’s graphics are pretty impressive, too. It’s not a statistical analysis, but it’s suggestive of a relationship between climate change and fire size/destructiveness.
The issue of precipitation is more complex than looking at total summer rain vs. acres burned, as others have pointed out. Winter precipitation is important, either for drought or for fuel load. Long-term drought is also a potential factor. So it’s no surprise that part of the analysis wasn’t significant. Willlis is right that he found no evidence for a link between precipitation and acreage burned, but that doesn’t mean there isn’t a link.
More interesting things on acres burned
Willis you gave the source but looking at that source I am wonder what figures you took from the documents
Only cal fire? or other agencies as well?
Total acres burned, all agencies.
w.
All US sceptics should read the NYT’s updates on climate change abd its impact on California’s inceasing numbers and more extreme wildfires. As more events like the Camp fire occur, the more the mainstream American public will require action from their government to reduce co2 emissions. Less CO2 = less extreme wildfires = less acreage burnt.
https://www.nytimes.com/2018/11/14/climate/california-wildfires-and-reinventing-air-conditioners-climate-newsletter.html?emc=edit_clim_20181114&nl=climate-fwd&nlid=7847975920181114&te=1
RE Ivan:
**As more events like the Camp fire occur, the more the mainstream American public will require action from their government to reduce co2 emissions. Less CO2 = less extreme wildfires = less acreage burnt.**
Sure, all skeptics AND CAGW types should look up and find me proof that more CO2 gives more extreme wildfires. As well all CAGW types should find me a scientific study that proves CO2 causes global warming aka climate change.