California Drought

Despite recent rains, California is still in the midst of crippling drought. In a recent opinion piece by Martin Hoerling the case is made that essentially, man-made climate change has nothing at all to do with the present California drought.

Thus, the scientific evidence does not support an argument that human-induced climate change has played any appreciable role in the current California drought.

I disagree.

Martin Hoerling is a smart guy, and I wouldn’t call him a “denier.” But we all know how easy it is for smart people to be stupid. It’s even possible for those who aren’t deniers, to deny. I think that’s exactly what Martin Hoerling has done.


There’s a lot of huffing and puffing in the article, but his argument basically boils down to this:

We can also say with high confidence that no appreciable trend toward either wetter or drier conditions has been observed for statewide average precipitation since 1895.

The ironic part is that Hoerling makes this the centerpiece of his argument in spite of emphasizing that drought is not just about precipitation.

It’s true that there’s no significant trend in California precipitation since 1895. Here’s precipitation anomaly, which reduces the counfounding influence of the annual cycle (it tends to be wetter during winter, especially January):


The estimate trend from linear regression (shown by the thick blue line) is decreasing, but not statistically significant even if you don’t correct for autocorrelation. If you do correct for autocorrelation, it’s definitely not statistically significant.

But as Hoerling himself states, drought isn’t just about precipitation. If we look at an actual measure of drought — the Palmer Drought Severity Index, or PDSI — then there is a decreasing trend (which means, toward more and/or more extreme drought) which is statistically significant, even after correcting for autocorrelation:


The blue line shows the trend estimated by linear regression, the red line shows a lowess smooth. Yes, there’s a trend toward more drought in California. Yes, it’s statistically significant. Martin Hoerling’s claim that there’s “no appreciable trend … for statewide average precipitation” is true, but misleading. What’s worse, he really ought to know better.

A similar, but weaker, trend shows up in a common measure of short-term drought, the Palmer Z-index:


Note that this last January (the last month for which data are presently available from the National Climate Data Center) is the lowest on record. We can see the trend more clearly if we just plot the smoothed version, together with the estimated linear trend.


After correcting for autocorrelation, the linear trend is not quite significant at the 95% confidence level. It’s “only” significant at the 94.8% confidence level. Statistically speaking, it’s one of those things that make you go “Hmmm…”

But instead of making Martin Hoerling go “Hmmm…” it seems to have made him go wrong. Or perhaps he hasn’t even looked at the PDSI or Z-index — in which case, what the hell is he doing pontificating about the California drought? He tries his best to assert that it’s definitely not the case that “human-induced climate change has played any appreciable role in the current California drought.” It sure reads to me as though he isn’t saying “Hold on, let’s not be so sure,” like a genuine skeptic — instead he’s asserting that he knows it ain’t so — like a denier.

What’s worse, for Californians, is that the snowpack in the Sierra Nevada mountains is so low that one can expect drought conditions to persist, possibly even worsen, throughout the year. An extreme example (yes it’s cherry-picked, in order to make a point) is the data from the SNOTEL site at Independence Creek:


Each line represents the snow-water-equivalent (SWE) in the snowpack throughout a single year. The thick red line is the SWE for this year. Notice anything striking about it? Here’s a close-up of the first three months, to make this year’s snowpack easier to compare to previous years’:


Given the current state of affairs, the outlook for California’s water supply this year isn’t just “not good,” it’s grim.

Maybe Martin Hoerling thinks he’s being the “honest broker” and arguing for some sort of “restraint” in assigning a role of man-made global warming. If he thinks that, he’s fooling himself, because he really doesn’t give the impression “hold on, let’s not be so sure,” he gives me the impression “there’s no connection at all and I’m sure of it.” And in that, he’s just plain wrong.

In my opinion, he’s also being irresponsible. Tremendously irresponsible. Man-made climate change is a huge threat to California, and this year’s drought is a perfect example why. It should be an alarm to “Wake up and smell the coffee,” but Martin Hoerling is saying “There is no coffee.”

One of the saddest aspects of this is that the present situation was predicted. The prediction was stunningly accurate — but I do realize that it’s possible to make a prediction which is stunningly accurate because you got lucky. Nonetheless, a stunningly accurate prediction is powerful evidence, and if correct, it means that man-made global warming was actually the cause of the present California drought, because man-made global warming certainly did cause the disappearance of Arctic sea ice. According to the scientists who made the prediction, prospects for the future are not good. I’d go so far as to say “grim.”

45 responses to “California Drought

  1. Michael Sweet


    In your graph of the PDSI, the Lowess curve appears to change slope around 1990. Can any interpretation be assigned to such a change or is it too early to determine the significance? How long would the new slope have to continue to determine the change was significant? If it becomes significant, what does that mean?

    Your stuff is always interesting. I understand a lot more about what I don’t know about statistics ;).

    [Response: Good questions, which deserve closer study before answering.]

  2. That should be Martin, or Marty, Hoerling, not Matt. Maybe you are thinking of Matt Groening!

    [Response: Thanks, fixed.]

  3. Ralph Snyder

    I see your argument that drought in California has gotten worse, and I certainly agree that climate change is a problem for California, but I do not see that you have established any connection between increasing drought andclimate change. (It would be surprising if it were not) Did I miss something?

    [Response: The two principal factors in drought are precipitation and temperature (leading to evaporation, and influencing the timing of snowmelt). If precipitation hasn’t trended but drought has, the logical cause is temperature (which has increased). That’s not proof — but it does show the folly of asserting with confidence that man-made climate change has no connection.]

  4. Harold Brooks

    Marty, not Matt, Hoerling

    [Response: Thanks, fixed.]

  5. Robert Damon

    I read Matt Hoerling’s piece this morning in the Times. I wondered how long it would take for someone to call him out on this, particularly in light of Holdren’s piece in response to Pielke, Jr. Happy (but not surprised) to see your response.

  6. Thanks so much for this nice debunking.of Hoerling. We should not let the NYTimes off the hook. Editors enable and encourage this and share culpability. Putting it in the Opinion pages is not responsible journalism.

    Thanks for showing them how it should be done.

    • John Garland

      Disagree somewhat: Putting it in the “journalistic opinion” pages is much more correct than calling it “objective news”.

  7. You nailed this on multiple levels — great post!

  8. I was about to make a comment on it, but Tamino has already made it:
    If the temperature goes up, but rain/precipitations stay the same, increased evaporation due to the increase in temp will make any drought to be worse, to have worse effects than previously.

  9. climatehawk1

    Thanks for addressing this so quickly.

  10. The Deniers are going to mention changes in water usage for agriculture to further confuse any specific links with climate and drought. I fear this could be very effective for decades to come. Its very hard to clearly link increases in temperature and evaporative rate to drought. On the other hand it seems totally illogical as to how increased temperatures wouldn’t increase drought severity or persistence.

  11. Yes, also thanks from me. Snow is a form of precipitation right ?

    • Yes.. but even more so, Snow is a way of storing winter precipitation for summer use. If you get less snow, or the precipitation falls as rain which can’t all be stored, then you lose water overall. It’s damm hard to stop every river..

  12. Tamino, I think that you should submit this to NYT as an OpEd rebuttal. If they don’t publish it, someone else will.

  13. Nice post. It is not just CA, drought has been on the increase over the western USA

    and the southwestern USA,

    Ironically, the data in the above links are from NOAA, the same agency that Hoerling works for.

    I’m curious what the independent SPEI data show for CA. Having briefly looked at the data over an area that roughly bounds CA, the same increase in drying is evident.

  14. Is there such a thing as a Pielke Proximity Index?

  15. Thanks for another timely article, Tamino. It’s one thing for Martin Hoerling to say that he can’t separate influences. It’s quite another thing for him to say that global warming categorically *hasn’t* influenced (insert whatever weather here).

    Martin Hoerling is known for this sort of thing.

  16. Part of his argument seems to be that while he thinks the drought isn’t particularly outstanding, the demand for water in the area (and thus the impact felt) is.

    A sort of Pielke Juniorism.

    I’ve always thought that if we’re making ourselves increasingly vulnerable to climate changes (and even quite small ones), then that’s all the more reason to want to mitigate quite quickly… no?

    Also… Have you contacted Hoerling for a response?

  17. I remember he previously attacked James Hansen’s work in the same area for a similar reason, back in 2012.

  18. 20 years on from the first IPCC report and on an anecdotal level the weather has caught up with climate change- here in the UK the wettest winter on record [despite one of those 3 months being dryer] we now bask in summer sunshine [in March]. So the caveat that no weather event can be attributed to AGW is on a non science level rather irrelevant.

    One swallow does not make a summer- or 3 100 year storms in one winter make AGW- but as much as I understand the need for caution from science I look to the sports analogy and drug taking- it is now like asking which races did Lance Armstrong actually win despite the steroids?

  19. Pete Dunkelberg

    Global warming and changes in drought.
    Kevin E. Trenberth, Aiguo Dai, Gerard van der Schrier Philip D. Jones Jonathan Barichivich Keith R. Briffa, Justin Sheffield 2013

    Note that Dai and Sheffield are co-authors. They seemed slightly at odds in preceeding papers. Here they iron it out.

    General drought factors:
    Drought or not = precipitation – (runoff + evaporation)

    runoff may increase due to more of the same annual precipitation occurring in slightly fewer but harder rains, and may be tracked via stream flow.

    When the air is dry there is increased evaporation from the ground if the temperature is higher. But soon the dry ground has no more evaporative cooling to give so the region gets even warmer.

    Depth of water table is another source of data.

    btw the linked pdf is unfriendly to copy & paste. Each word becomes a line. Is there a solution for that?

  20. Pete Dunkelberg

    Tamino: “In my opinion, he’s also being irresponsible.”
    Evidence: His by line indicates he is an expert yet he took his material from Revkin. And seems not to have checked the PDSI.

  21. Tamino,

    Another fascinating post, and I agree completely that Hoerling is overstepping on his conclusions.

    I still think I’m missing something here, though. First, you agree there is no significant trend in precipitation, and then – quite correctly – argue PDSI is a better measure for droughts. This made perfect sense until I read the article referenced in this sentence: “One of the saddest aspects of this is that the present situation was predicted.” All of a sudden, it seems the focus shifts back to precipitation alone.

    From the story (and the associated abstract – the full article is too expensive for me), it appears that these scientists’ prediction was entirely based on precipitation: “They used powerful computers ‘to simulate the effects of reduced Arctic sea ice,’ and ‘their most striking finding was a significant reduction in rain and snowfall in the American West.'”

    Although California’s current situation looks stunningly similar to their prediction (and as Hoerling states, this year’s drought is associated with very low precipitation), one year doesn’t make a trend (as is obvious with last year’s minimum sea ice extent, for example). As you show, we haven’t seen a “significant reduction in rain and snowfall in the American West” – at least on climate time scales – so I would hesitate to argue that their prediction has come true (yet).

    The context of your “predicted” sentence seems to indicate that acting immediately could have reduced the impact of the drought (and I’m in full agreement that we are way past the time to act on AGW), but precipitation trends from the time the article was published until last year would not have supported their hypothesis any more than one could claim that this year alone vindicates it.

    I may be completely wrong (the full article may be making the case based on PDSI, not precipitation), and if so, I would greatly appreciate enlightenment from someone with access to the article.

    • I was a little puzzled by this too. I think it could have been written more clearly, but after a little thought I came to the conclusion that that they’re talking about slightly different things. The PDSI was used to show there is an observable (if not strictly statistically significant by the usual measure) trend in drought for the region. The PDSI being a better measure of drought that total precipitation for reasons explained in Tamino’s post.

      This seems to be a separate (although obviously related) point to the prediction that increased blocking patterns would cause more frequent or more severe droughts, the current situation being an example of this. The fact that there isn’t an observable trend in the precepitation data suggests that this predicted increase is not yet supported by the evidence. However, I think it would be wise to take a closer look at the precipitation patterns before reaching this conclusion. It may be that these blocking patterns are in fact occuring more regularly and resulting in periods of reduced precipitation, but these are then ‘balanced out’ in the total precipitation data by periods of heavier rainfall in between. This would still lead to increased drought without necessarily leaving a trend in the aggregated precipitation data.

  22. Hoerling states that that the drought “has occurred principally because of a lack of rain, not principally because of warmer temperatures.” If the Palmer index and related Z-index have a term in the algorithm for temperature and therefore related evapotranspiration, how does Hoerling “know” the assertion about temperature is true, if the two indices are showing the anomalies they are showing? That mountain snowmelt is occurring earlier is well-known, for example, due to earlier establishment of a dome of high pressure over the entire western US.

  23. I’d like to get a response on this from a poster on another blog.

    “I am afraid Tamino is the one who isn’t being an honest broker here. What he has done is taken the Palmer drought index and run a regression through it….this does not tell you if there has been more drought necessarily rather that there is an overall trend albeit slight to a higher severity index. And that trend is seriously impacted by the very unusual decade of very wet conditions from 1905 to 1915.

    If you look at the actual number of years of drought there are several periods in the past with drought conditions as long or longer (as the case in the long lasting eighties California drought) and the most severe was actually near the turn of the century just prior to the extended period of wet conditions.”

    [Response: If you read carefully you’ll note that I said “toward more and/or more extreme drought.” How far in denial do you have to be to protest that I’ve *only* shown “trend albeit slight to a higher severity index”? I did later say “more drought,” so those who want to pick nits have an excuse to support their denial.

    As for “seriously impacted by the very unusual decade of very wet conditions from 1905 to 1915,” if you just use the PDSI data after 1915 there’s still a downward trend and it’s still statistically signficant. What makes that argument doubly pathetic is that you can always erase a trend you don’t like if you exclude the data you don’t like — he just didn’t go far enough. Maybe he needs some cherry-picking lessons.

    It also seems to miss the point of this post, which is not to prove a definitive connection between man-made climate change and the California drought, but to show the folly of Hoerling’s implication that there’s definitely no connection. Those who will argue that the connection is uncertain, make my point for me.]

  24. @Gary,
    Any measure, PSDI or not, will have imperfections. P-values have imperfections. I bet, however, that if a frequestist bootstrap on the slope of the linear regression which Tamino did is done, you’ll find strong evidence that there’s a pony here, not something which can be dismissed by deleting data to shove it over a particular P-value. I bet also that if a Bayesian regression were done on the same data using PSDI, you’d see probability mass in the resulting posterior of the slope which would make the case. I could try that, but I don’t know where to get the PSDI data or the drought data. Was it given some place above? Or in Hoerling’s op-ed?

    [Response: It’s here.]

  25. I hope some people are prepared to pay more for fruit and vegetables, since CA grows about half of them for the US.

    Hoerling manages to ignore one of the clearest impacts of higher temperatures, an issue shared with Colorado and others, although I don’t know their numbers offhand.

    There were some allusions to this, but to be more explicit:
    1) About half of CA’s reservoir capacity is Sierra snowpack.
    2) We could get the same total precipitation as decades before, but if less of it falls as snow and/or the snow melts faster:
    a) the Central Valley gets massive floods in the Spring
    b) and then, there is less water in summer or fall.

  26. Thanks, I actually already made a similar comment in rebuttal, but I wanted to pass something along from the horse’s mouth for a change. Oh, and Rockdoc, in case you’re reading this feel free to jump in here. I’m sure the blog would appreciate the perspective from a petroleum geologist’s point of view.

  27. OT–what’s up with RC? All of a sudden I’m getting ‘Forbidden’ error message when I try to go there. Anybody know what’s going on?

    [Response: I got the same message. I hope it’s only brief.]

  28. Horatio Algeranon

    “Debate and switch”
    — by Horatio Algeranon

    Debate and switch is all the rage
    Present stands for future age
    Cause of current drought’s in doubt
    Which means that climate change is out

  29. Francis Cels

    How long must this drought go on before we act?

  30. San Francisco radio mentioned yesterday that SF residents, asked to try to cut water use by 10 percent, have cut water use by 18 percent.

  31. Oops. Looks like humanity is surprisingly similar to an asteroid strike:

    ” a collision with sulfate rocks at a velocity of 15 to 20 kilometers per second. That type of stone can be found in geological layers around the Chicxulub crater on the Yucatan Peninsula, the believed site of the impact.
    The team examined the gas generated by the collision and discovered the vapor mainly consisted of sulfur trioxide, a causative substance of acid rain. Sulfur trioxide easily binds to rain droplets.
    Ono and his colleagues estimated that more than 100 billion tons of sulfur trioxide covered the planet after the asteroid strike, and stronger acidic rain than gastric juices in the stomach fell to the ground within several days following the impact. The oceans are believed to have remained acidified for more than a year, to the extent where most of the plankton, a vital source in the food chain, could not survive.”

  32. In the past, the Central Valley of California was very wet in the spring, and that water evaporated and became the last layer of the Sierra Snow pack. Over the last 50 years we have drained most of those ephemeral wetlands, so the water goes to the ocean instead of sitting in Central Valley until it can warm and evaporate.

    The water in the CV ephemeral wetlands,was a large stock of water that was included in water rights. It was a valuable resource that we simply flushed down the drain.

    Our draining of the CV is an additional cause of the ongoing drought.

  33. The West Without Water is an effort to explain how variable the California climate has been in the last several thousand years, i.e. how variable it really is. Droughts have been far more severe and have lasted a lot longer than anything seen since California became a state.

    • Yes, that’s well-known to anyone who’s read anything on paleoclimate–mentioned in “6 Degrees,” “Hockey Stick,” “Fixing Climate,” and maybe “Long Thaw”, for good measure. (I’ve published summaries of all of these.)

      However, it’s also true that a Southwest much drier than today’s is a robust result of model projections under increasing CO2 burdens.

    • Just ask the Anasazi. It didn’t end well.