How much is a “big change” in the climate?

Here’s the density of temperature changes over centuries. I used these data and calculate the average change in temperature per century.
ice_core_temp_changes
Last century’s temperature increase of 0.8 degrees C was an outlier (but not an extreme outlier). About 95% of temperature changes were slower than last century’s temperature change. If the climate models are correct and the world sees a 2.5 degree increase, this would be an extreme outlier. Only about 1 in a three or four hundred centuries sees that dramatic of temperature changes.

This, of course, doesn’t guarantee catastrophe, but it suggests we should at least insure ourselves against the possibility of catastrophe.

11 Responses to “How much is a “big change” in the climate?”

  • Kevin Dick says:

    Given what you just stated, wouldn’t you therefore conclude that the climate models are likely to be wrong? After all, they didn’t predict the current century’s temperature increase a priori. Therefore, you would expect some mean reversion. In fact, we’ve seen a great deal of mean reversion since 2000. It might be interesting to do the same thing for decadal temperature differences then look at the different decades in the 20th century as well as the 10 years from 1999-2009.

  • swong says:

    Couldn’t you also apply that logic to, say, economic growth in Los Angeles county from 20000 BC to now? Zero, zero, zero, zero, zero, zero, little blip, major spike in the 20th century, must be an outlier, growth in the 21st century is improbable…

  • Kevin Dick says:

    There was that little part about a priori prediction that is the crucial difference. You know, scientific method and all that.

    There are plenty of economic models that have demonstrated some skill in predicting LA economic growth on yearly or decadal time scales. Within the bounds of that skill, you should absolutely take into account those predictions when making decadal-scale plans that depend on economic growth in LA.

    The problem with climate models is they have demonstrated no forecasting skill on any time horizon, least of all centuries.

  • swong says:

    Ah sorry, I meant to address the comment to Will’s original post. Should have been clearer about that.

    Re: Year 2000 – we hit the peak of the solar cycle in 2000, and are at or around the solar minimum right now.

  • pushmedia1 says:

    swong, your comment was a better critique of Kevin’s comment then my post!

    The point of my post was that if the climate models are right then we’re headed for relatively uncharted territory.

    I should note that for much of the data I estimated century changes using data that spans more than centuries. For example, the ice cores can tell us that the difference in temperatures between 15,000 and 15,500 years ago, but not the centuries in between. I divide that difference by five to get the average change by century. This means that actual changes in temperature can be more volatile than what I reported above. Suppose between 15,500 and 15,400 years ago temperatures increased by 10 degrees and then in the next 400 years, they decreased by 2 degrees a century. The data would tell us that the average temp change was less than a degree a century but the actual volatility was much higher. Anyway, the bottom line is that the “1 in 3 or 4 hundred centuries” estimate is a lower bound. I don’t know how to estimate the size of this bias.

  • swong says:

    On a reread of your post, it looks like you’re just pointing out that deltas this big are pretty rare; nothing more or less. If I’m not mistaken, the Younger Dryas period saw some major temperature deltas on a decade scale – I’ll have to dig up some sources for that though.

    Re: accuracy of climate models – my understanding is that while their forecasts aren’t spot on, their margin of error isn’t all that wide when applied to the climate record either. On the other hand, the recent rate of arctic ice pack melting is actually above the upper bound of what the best models predicted.

  • Kevin Dick says:

    If you want to read a formal indictment of climate models by a forecasting expert, see here:

    http://repository.upenn.edu/cgi/viewcontent.cgi?article=1144&context=marketing_papers

    As for Arctic ice melting, that proves my point. Mean surface temperature are below those predicted by the model, Artic melting is above, Antartic _formation_ is above. Basically, they can’t predict any interesting feature of the climate system even a decade in advance.

  • swong says:

    Hm. I’m not a climatologist or an economist, but that paper looks a little funky. I see a lot of references to readability and scans for forecasting references, and not really any actual, direct refutations of any claims. The Gunning fog index helps argue that someone’s writing is inaccessible, but it doesn’t say anything about their conclusions.

    Do you mean Antarctic ice formation? Area or volume?

  • Kevin Dick says:

    That paper was making the point that climate “forecasts” are not really forecasts in the professional sense of the word and that the documents are constructed in such a way as to obscure that fact.

    If you want the scientific refutation of claims try: http://www.pas.rochester.edu/~douglass/papers/Published%20JOC1651.pdf

    For Antarctic ice, I was thinking of both sea ice extent and ice sheet mass:

    Extent: http://arctic.atmos.uiuc.edu/cryosphere/IMAGES/current.anom.south.jpg

    Mass: http://www.cpom.org/research/djw-ptrsa364.pdf

  • ian fairchild says:

    First, to Will’s post, this is a nice exhibit. There is new information for me. I’m a proud denier, but like to see well presented data from the other side. Anyway, two things to comment. The models suckkkk. Take the predicted temperature from the model and plot it against the exhibit. Should be interesting.

  • pushmedia1 says:

    ian, the comment in my post is conditional on the models being right. Personally, I don’t have much faith in the models, either. I analogize them to macro models, which I do know something about. Macro models are just ok at forecasting a few quarters in advance.

    But a lot of people that know a lot more than I do about this stuff seem to take the models more seriously. This forces me to take them more seriously than I otherwise would.