Friday, March 6, 2009

Hume-an Error

I didn't intend for this to be a climate blog. It was just supposed to be a place where I could note odd stuff that I run across and to let off steam about foolishness that I see.

But people will keep spouting the most ignorant crap about global warming, so here we go again.

The latest loony tirade comes from the mouth of the redoubtable Brit Hume (incidentally, I think this argues for an entirely new definition of redoubtable: someone you can doubt over and over again.) Talking about a recent demonstration in Washington, he said this on the March 2 edition of Fox News's Special Report:
[Y]ou have to give those global warming activists credit for pluck. Not only were they protesting warming temperatures in a city going through its coldest winter in recent memories—a city in the midst of a snow emergency and sub-freezing temperatures—they were also doing so on a planet that has seen no average warming for the past 10 years. But climate change alarmists are not easily fazed.
...
The problem with [scientists' climate models] is that when data from the past have been plugged into them, they have had trouble predicting today's temperatures. The climate alarmists certainly did not foresee the cooling trend of the past decade. No matter.
Let's skip over whether or not he should be using the loaded term "climate change alarmists" twice in something that they're calling a "report." I guess that's why Fox News's slogans are "Unfair and Unbalanced" and "We Spin, You Listen Up." (I got those right, didn't I?)

While we're at it, we can also skip the bit where he apparently just makes up the "coldest winter in recent memories" factoid: a full third of the previous nine DC winters were colder.1 Maybe he has a really bad memory.

Nah. He just likes to make stuff up.

OK, that's enough Andy Rooney. Let's get to the serious issues.

"Coldest winter in recent memories"

Even if this were true—and it's not—this is a classic case of confusing weather and climate. Weather is something that happens day-to-day, month-to-month, year-to-year. There's a lot of variability—noise, if you will—in it. It's very hard to predict more than a couple of days in advance. Climate is, in essence, weather with the noise removed.

Climate information is obtained by averaging weather over a long period of time and observing trends. Suppose you were to graph annual temperatures over many years. If you observe the temperature trends and how they are changing, you are looking at climate; but if you observe the temperature spikes here and there, you are looking at weather.

This winter's average temperature is weather, not climate. (Whether or not there's a snowstorm that discourages global warming alarmists from demonstrating is really weather.)

Let's use a classic example: rolling dice. If you roll an ordinary six-sided die many times, you will find that it averages around 3.5. If you graph this, you'll find a lot of spikes for low and high rolls, but if you make a trend line, it will be more or less flat. Here's a little graph of a hundred simulated throws:



The trend line is in red. That is climate. The line is essentially flat because the dice rolls aren't trending up or down—there's no change in the average roll as we move along in time.

The actual throws are in blue. That is weather. These vary rather wildly. Some throws are near the trend line, some are way above it, and some are way below it.

Now here is another simulation. This time I've loaded the dice: as we go on, it gets progressively easier to roll higher numbers, so now the average roll is increasing as we move along in time (this simulates global warming):



The loading of the dice is clearly visible in the upward curve of the trend line. But notice where we rolled a 2 at the red arrow. It doesn't prove that the dice aren't loaded, right? In fact, we can't tell anything at all from the one data point.

If this were a graph of average DC winter temperatures instead of dice rolls, the arrow would point to our "cold winter" (and, yes, even with global warming, there will be relatively cold winters). Just as that one throw can't tell us anything about whether the dice are loaded, that one cold winter can't tell us anything about global warming.

Here's the key point: weather is, for all intents and purposes, random. Climate is not. And you can't look at that random weather for today or this month or this year and use it to say anything about climate.2 You can't just look at a single point in time and say, "It's cold, therefore global warming is bunk." But that's what Brit did with his "it's cold this winter" comment. (Maybe he'll come back in August when it's 105° in DC and say, "It's really hot today. Looks like I was wrong about global warming.")


"Cooling trend of the last decade"

Brit simply asserts this, so we don't know where he got it from—but I think I can make a pretty good guess. A lot of people (including George Will, quite recently—coincidence?) have been saying the same thing, and it always seems to come down to this: it was a little cooler in 2008 than it was in 1998.

OK, let's go back to our loaded dice. Look at the green line I've added:



See that, loaded dice alarmists? The recent trend is downward!

Well, not really. I just drew a pretty line between two arbitrary rolls. It doesn't mean anything at all.

But that green line is what makes Brit (or whoever he got this nonsense from) say that there's been no warming for the last ten years: he picked two arbitrary years, drew a line between them, and said, "See? No warming!" Unfortunately for Brit, it doesn't work that way. The individual data points are random, and you can't draw any conclusions by comparing two random things. Just as the green line here doesn't show that the dice aren't loaded (because they are loaded), the fact that 2008 was a little cooler than 1998 doesn't mean that there's no climate warming going on (because climate warming is going on).

(By the way, 2008 was cooler than 1998 in large part because 1998 was an El Niño year, while 2008 was a La Niña year—El Niño has a warming effect, while La Niña has a cooling effect. But despite La Niña, 2008 was still the 10th warmest year on record. There's more about this claim in the George Will response.)

"They have had trouble predicting today's temperatures"

Well, this one is real easy. "Today's temperatures" is weather. You can't predict weather from climate models. Repeat after me, Brit: for all practical purposes, weather is random. Climate models do not try to predict weather. You can't predict today's weather—or this year's weather, for that matter—from any climate model. That's not what they're for. Climate models try to predict the red line, not the blue line.

And the climate models are, in fact, rather good at doing that. Scientists have gone back to look at some of the older models and have found that longer term temperature trends have been pretty much as expected. RealClimate (a great site run by actual climate scientists) has more information.

Brit, you're supposed to be a journalist. You got some splainin' to do.


1 Source: Weather Underground. The mean winter (December 1-February 28) temperatures for Washington, DC. in 2000, 2002, and 2003 were all lower than 2008.

2 Even a decade is a bit dicey (sorry). For global surface temperatures, according to climate scientist Gavin Schmidt, 15 years is the point at which weather "noise" averages out. What this means is that, if the climate were not changing at all, the average temperature for any 15-year span would be about the same as that for any other 15-year span because weather averages out over a period that long. That is not be true for a shorter span such as a decade—two different decades could have significantly different average temperatures even with no climate change. So, 15 years is about the shortest time span you can use to say anything really meaningful about global climate trends.

Unless you account for noise.

If you do that, you can point to a shorter time period as being anomalously warm or cool. The weather noise for a decade has been calculated, and it's less that 0.1°C. So, in the absence of climate change, we would expect the average temperature for any decade to be within 0.1°C of the long-term average. If the average for a particular decade is more than 0.1°C different from the long-term average, we can say that it's an anomaly and possible evidence of a change in the climate.

3 comments:

  1. Nicely written, Chris. Let's hope repsonses like this start to add up and get the message through against the firehose of repetition of these really lame, badly reasoned statements.

    - Jim

    ReplyDelete
  2. Very good,

    But I worry that you're preaching to the choir.

    Surely Hume et. al. don't really think that if one year is cold, global warming is false, right? Are they really THAT stupid?

    I hope there's a way to get this information to the (right-wing) "listeners", because obviously the (right-wing) "speakers" won't lift a finger to educate their masses.

    ReplyDelete
  3. Thanks, Frank (and bbs). We can only keep trying.

    ReplyDelete